fix purego build errors on non-supported architectures

* NVML: only read memory usage if utilization > 0% to allow rtd3, #1522

* logging: /x/exp/slog -> log/slog everywhere, fixes log instance inconsistencies

.dockerignore

@@ -0,0 +1,48 @@
+# Node.js dependencies
+node_modules
+internalsite/node_modules
+
+# Go build artifacts and binaries
+build
+dist
+*.exe
+beszel-agent
+beszel_data*
+pb_data
+data
+temp
+
+# Development and IDE files
+.vscode
+.idea*
+*.swc
+__debug_*
+
+# Git and version control
+.git
+.gitignore
+
+# Documentation and supplemental files
+*.md
+supplemental
+freebsd-port
+
+# Test files (exclude from production builds)
+*_test.go
+coverage
+
+# Docker files
+dockerfile_*
+
+# Temporary files
+*.tmp
+*.bak
+*.log
+
+# OS specific files
+.DS_Store
+Thumbs.db
+
+# .NET build artifacts
+agent/lhm/obj
+agent/lhm/bin

.github/workflows/docker-images.yml

@@ -10,51 +10,141 @@ jobs:
     runs-on: ubuntu-latest
     strategy:
       fail-fast: false
+      max-parallel: 5
       matrix:
         include:
+          # henrygd/beszel
           - image: henrygd/beszel
-            context: ./beszel
-            dockerfile: ./beszel/dockerfile_hub
+            dockerfile: ./internal/dockerfile_hub
             registry: docker.io
             username_secret: DOCKERHUB_USERNAME
             password_secret: DOCKERHUB_TOKEN
-
+            tags: |
+              type=raw,value=edge
+              type=semver,pattern={{version}}
+              type=semver,pattern={{major}}.{{minor}}
+              type=semver,pattern={{major}}
+              type=raw,value={{sha}},enable=${{ github.ref_type != 'tag' }}
+            
+          # henrygd/beszel-agent:alpine
           - image: henrygd/beszel-agent
-            context: ./beszel
-            dockerfile: ./beszel/dockerfile_agent
+            dockerfile: ./internal/dockerfile_agent_alpine
             registry: docker.io
             username_secret: DOCKERHUB_USERNAME
             password_secret: DOCKERHUB_TOKEN
+            tags: |
+              type=raw,value=alpine
+              type=semver,pattern={{version}}-alpine
+              type=semver,pattern={{major}}.{{minor}}-alpine
+              type=semver,pattern={{major}}-alpine
 
+          # henrygd/beszel-agent-nvidia
           - image: henrygd/beszel-agent-nvidia
-            context: ./beszel
-            dockerfile: ./beszel/dockerfile_agent_nvidia
+            dockerfile: ./internal/dockerfile_agent_nvidia
             platforms: linux/amd64
             registry: docker.io
             username_secret: DOCKERHUB_USERNAME
             password_secret: DOCKERHUB_TOKEN
+            tags: |
+              type=raw,value=edge
+              type=semver,pattern={{version}}
+              type=semver,pattern={{major}}.{{minor}}
+              type=semver,pattern={{major}}
+              type=raw,value={{sha}},enable=${{ github.ref_type != 'tag' }}
 
+          # henrygd/beszel-agent-intel
+          - image: henrygd/beszel-agent-intel
+            dockerfile: ./internal/dockerfile_agent_intel
+            platforms: linux/amd64
+            registry: docker.io
+            username_secret: DOCKERHUB_USERNAME
+            password_secret: DOCKERHUB_TOKEN
+            tags: |
+              type=raw,value=edge
+              type=semver,pattern={{version}}
+              type=semver,pattern={{major}}.{{minor}}
+              type=semver,pattern={{major}}
+              type=raw,value={{sha}},enable=${{ github.ref_type != 'tag' }}
+
+          # ghcr.io/henrygd/beszel
           - image: ghcr.io/${{ github.repository }}/beszel
-            context: ./beszel
-            dockerfile: ./beszel/dockerfile_hub
+            dockerfile: ./internal/dockerfile_hub
             registry: ghcr.io
             username: ${{ github.actor }}
             password_secret: GITHUB_TOKEN
+            tags: |
+              type=raw,value=edge
+              type=semver,pattern={{version}}
+              type=semver,pattern={{major}}.{{minor}}
+              type=semver,pattern={{major}}
+              type=raw,value={{sha}},enable=${{ github.ref_type != 'tag' }}
 
+          # ghcr.io/henrygd/beszel-agent
           - image: ghcr.io/${{ github.repository }}/beszel-agent
-            context: ./beszel
-            dockerfile: ./beszel/dockerfile_agent
+            dockerfile: ./internal/dockerfile_agent
             registry: ghcr.io
             username: ${{ github.actor }}
             password_secret: GITHUB_TOKEN
+            tags: |
+              type=raw,value=edge
+              type=raw,value=latest
+              type=semver,pattern={{version}}
+              type=semver,pattern={{major}}.{{minor}}
+              type=semver,pattern={{major}}
+              type=raw,value={{sha}},enable=${{ github.ref_type != 'tag' }}
 
+          # ghcr.io/henrygd/beszel-agent-nvidia
           - image: ghcr.io/${{ github.repository }}/beszel-agent-nvidia
-            context: ./beszel
-            dockerfile: ./beszel/dockerfile_agent_nvidia
+            dockerfile: ./internal/dockerfile_agent_nvidia
             platforms: linux/amd64
             registry: ghcr.io
             username: ${{ github.actor }}
             password_secret: GITHUB_TOKEN
+            tags: |
+              type=raw,value=edge
+              type=semver,pattern={{version}}
+              type=semver,pattern={{major}}.{{minor}}
+              type=semver,pattern={{major}}
+              type=raw,value={{sha}},enable=${{ github.ref_type != 'tag' }}
+
+          # ghcr.io/henrygd/beszel-agent-intel
+          - image: ghcr.io/${{ github.repository }}/beszel-agent-intel
+            dockerfile: ./internal/dockerfile_agent_intel
+            platforms: linux/amd64
+            registry: ghcr.io
+            username: ${{ github.actor }}
+            password_secret: GITHUB_TOKEN
+            tags: |
+              type=raw,value=edge
+              type=semver,pattern={{version}}
+              type=semver,pattern={{major}}.{{minor}}
+              type=semver,pattern={{major}}
+              type=raw,value={{sha}},enable=${{ github.ref_type != 'tag' }}
+
+          # ghcr.io/henrygd/beszel-agent:alpine
+          - image: ghcr.io/${{ github.repository }}/beszel-agent
+            dockerfile: ./internal/dockerfile_agent_alpine
+            registry: ghcr.io
+            username: ${{ github.actor }}
+            password_secret: GITHUB_TOKEN
+            tags: |
+              type=raw,value=alpine
+              type=semver,pattern={{version}}-alpine
+              type=semver,pattern={{major}}.{{minor}}-alpine
+              type=semver,pattern={{major}}-alpine
+
+          # henrygd/beszel-agent (keep at bottom so it gets built after :alpine and gets the latest tag)
+          - image: henrygd/beszel-agent
+            dockerfile: ./internal/dockerfile_agent
+            registry: docker.io
+            username_secret: DOCKERHUB_USERNAME
+            password_secret: DOCKERHUB_TOKEN
+            tags: |
+              type=raw,value=edge
+              type=semver,pattern={{version}}
+              type=semver,pattern={{major}}.{{minor}}
+              type=semver,pattern={{major}}
+              type=raw,value={{sha}},enable=${{ github.ref_type != 'tag' }}
 
     permissions:
       contents: read
@@ -68,10 +158,10 @@ jobs:
         uses: oven-sh/setup-bun@v2
 
       - name: Install dependencies
-        run: bun install --no-save --cwd ./beszel/site
+        run: bun install --no-save --cwd ./internal/site
 
       - name: Build site
-        run: bun run --cwd ./beszel/site build
+        run: bun run --cwd ./internal/site build
 
       - name: Set up QEMU
         uses: docker/setup-qemu-action@v3
@@ -84,17 +174,12 @@ jobs:
         uses: docker/metadata-action@v5
         with:
           images: ${{ matrix.image }}
-          tags: |
-            type=raw,value=edge
-            type=semver,pattern={{version}}
-            type=semver,pattern={{major}}.{{minor}}
-            type=semver,pattern={{major}}
-            type=raw,value={{sha}},enable=${{ github.ref_type != 'tag' }}
+          tags: ${{ matrix.tags }}
 
       # https://github.com/docker/login-action
       - name: Login to Docker Hub
-        env: 
-            password_secret_exists: ${{ secrets[matrix.password_secret] != '' && 'true' || 'false' }}
+        env:
+          password_secret_exists: ${{ secrets[matrix.password_secret] != '' && 'true' || 'false' }}
         if: github.event_name != 'pull_request' && env.password_secret_exists == 'true'
         uses: docker/login-action@v3
         with:
@@ -107,7 +192,7 @@ jobs:
       - name: Build and push Docker image
         uses: docker/build-push-action@v5
         with:
-          context: "${{ matrix.context }}"
+          context: ./
           file: ${{ matrix.dockerfile }}
           platforms: ${{ matrix.platforms || 'linux/amd64,linux/arm64,linux/arm/v7' }}
           push: ${{ github.ref_type == 'tag' && secrets[matrix.password_secret] != '' }}

.github/workflows/inactivity-actions.yml

@@ -10,12 +10,25 @@ permissions:
   pull-requests: write
 
 jobs:
+  lock-inactive:
+    name: Lock Inactive Issues
+    runs-on: ubuntu-24.04
+    steps:
+      - uses: klaasnicolaas/action-inactivity-lock@v1.1.3
+        id: lock
+        with:
+          days-inactive-issues: 14
+          lock-reason-issues: ""
+          # Action can not skip PRs, set it to 100 years to cover it.
+          days-inactive-prs: 36524
+          lock-reason-prs: ""
+
   close-stale:
     name: Close Stale Issues
     runs-on: ubuntu-24.04
     steps:
       - name: Close Stale Issues
-        uses: actions/stale@v9
+        uses: actions/stale@v10
         with:
           repo-token: ${{ secrets.GITHUB_TOKEN }}
 
@@ -32,6 +45,8 @@ jobs:
           # Timing
           days-before-issue-stale: 14
           days-before-issue-close: 7
+          # Action can not skip PRs, set it to 100 years to cover it.
+          days-before-pr-stale: 36524
 
           # Labels
           stale-issue-label: 'stale'

.github/workflows/release.yml

@@ -21,10 +21,10 @@ jobs:
         uses: oven-sh/setup-bun@v2
 
       - name: Install dependencies
-        run: bun install --no-save --cwd ./beszel/site
+        run: bun install --no-save --cwd ./internal/site
 
       - name: Build site
-        run: bun run --cwd ./beszel/site build
+        run: bun run --cwd ./internal/site build
 
       - name: Set up Go
         uses: actions/setup-go@v5
@@ -38,13 +38,13 @@ jobs:
 
       - name: Build .NET LHM executable for Windows sensors
         run: |
-          dotnet build -c Release ./beszel/internal/agent/lhm/beszel_lhm.csproj
+          dotnet build -c Release ./agent/lhm/beszel_lhm.csproj
         shell: bash
 
       - name: GoReleaser beszel
         uses: goreleaser/goreleaser-action@v6
         with:
-          workdir: ./beszel
+          workdir: ./
           distribution: goreleaser
           version: latest
           args: release --clean

.github/workflows/vulncheck.yml

@@ -15,7 +15,7 @@ permissions:
 
 jobs:
   vulncheck:
-    name: Analysis
+    name: VulnCheck
     runs-on: ubuntu-latest
     steps:
       - name: Check out code into the Go module directory
@@ -23,11 +23,11 @@ jobs:
       - name: Set up Go
         uses: actions/setup-go@v5
         with:
-          go-version: 1.24.x
-          cached: false
+          go-version: 1.25.x
+          # cached: false
       - name: Get official govulncheck
         run: go install golang.org/x/vuln/cmd/govulncheck@latest
         shell: bash
       - name: Run govulncheck
-        run: govulncheck -C ./beszel -show verbose ./...
+        run: govulncheck -show verbose ./...
         shell: bash

.gitignore

@@ -8,15 +8,15 @@ beszel_data
 beszel_data*
 dist
 *.exe
-beszel/cmd/hub/hub
-beszel/cmd/agent/agent
+internal/cmd/hub/hub
+internal/cmd/agent/agent
 node_modules
-beszel/build
+build
 *timestamp*
 .swc
-beszel/site/src/locales/**/*.ts
+internal/site/src/locales/**/*.ts
 *.bak
 __debug_*
-beszel/internal/agent/lhm/obj
-beszel/internal/agent/lhm/bin
+agent/lhm/obj
+agent/lhm/bin
 dockerfile_agent_dev

.goreleaser.yml

@@ -5,24 +5,36 @@ project_name: beszel
 before:
   hooks:
     - go mod tidy
+    - go generate -run fetchsmartctl ./agent
 
 builds:
   - id: beszel
     binary: beszel
-    main: cmd/hub/hub.go
+    main: internal/cmd/hub/hub.go
     env:
       - CGO_ENABLED=0
     goos:
       - linux
       - darwin
+      - windows
+      - freebsd
     goarch:
       - amd64
       - arm64
       - arm
+    ignore:
+      - goos: windows
+        goarch: arm64
+      - goos: windows
+        goarch: arm
+      - goos: freebsd
+        goarch: arm64
+      - goos: freebsd
+        goarch: arm
 
   - id: beszel-agent
     binary: beszel-agent
-    main: cmd/agent/agent.go
+    main: internal/cmd/agent/agent.go
     env:
       - CGO_ENABLED=0
     goos:
@@ -85,6 +97,9 @@ archives:
       {{ .Binary }}_
       {{- .Os }}_
       {{- .Arch }}
+    format_overrides:
+      - goos: windows
+        formats: [zip]
 
 nfpms:
   - id: beszel-agent
@@ -103,28 +118,28 @@ nfpms:
     formats:
       - deb
     contents:
-      - src: ../supplemental/debian/beszel-agent.service
+      - src: ./supplemental/debian/beszel-agent.service
         dst: lib/systemd/system/beszel-agent.service
         packager: deb
-      - src: ../supplemental/debian/copyright
+      - src: ./supplemental/debian/copyright
         dst: usr/share/doc/beszel-agent/copyright
         packager: deb
-      - src: ../supplemental/debian/lintian-overrides
+      - src: ./supplemental/debian/lintian-overrides
         dst: usr/share/lintian/overrides/beszel-agent
         packager: deb
     scripts:
-      postinstall: ../supplemental/debian/postinstall.sh
-      preremove: ../supplemental/debian/prerm.sh
-      postremove: ../supplemental/debian/postrm.sh
+      postinstall: ./supplemental/debian/postinstall.sh
+      preremove: ./supplemental/debian/prerm.sh
+      postremove: ./supplemental/debian/postrm.sh
     deb:
       predepends:
         - adduser
         - debconf
       scripts:
-        templates: ../supplemental/debian/templates
+        templates: ./supplemental/debian/templates
         # Currently broken due to a bug in goreleaser
         # https://github.com/goreleaser/goreleaser/issues/5487
-        #config: ../supplemental/debian/config.sh
+        #config: ./supplemental/debian/config.sh
 
 scoops:
   - ids: [beszel-agent]
@@ -135,7 +150,7 @@ scoops:
     homepage: "https://beszel.dev"
     description: "Agent for Beszel, a lightweight server monitoring platform."
     license: MIT
-    skip_upload: "{{ if .Env.IS_FORK }}true{{ else }}auto{{ end }}"
+    skip_upload: '{{ if eq (tolower .Env.IS_FORK) "true" }}true{{ else }}auto{{ end }}'
 
 # # Needs choco installed, so doesn't build on linux / default gh workflow :(
 # chocolateys:
@@ -169,7 +184,7 @@ brews:
     homepage: "https://beszel.dev"
     description: "Agent for Beszel, a lightweight server monitoring platform."
     license: MIT
-    skip_upload: "{{ if .Env.IS_FORK }}true{{ else }}auto{{ end }}"
+    skip_upload: '{{ if eq (tolower .Env.IS_FORK) "true" }}true{{ else }}auto{{ end }}'
     extra_install: |
       (bin/"beszel-agent-launcher").write <<~EOS
         #!/bin/bash
@@ -201,7 +216,7 @@ winget:
     release_notes_url: "https://github.com/henrygd/beszel/releases/tag/v{{ .Version }}"
     publisher_support_url: "https://github.com/henrygd/beszel/issues"
     short_description: "Agent for Beszel, a lightweight server monitoring platform."
-    skip_upload: "{{ if .Env.IS_FORK }}true{{ else }}auto{{ end }}"
+    skip_upload: '{{ if eq (tolower .Env.IS_FORK) "true" }}true{{ else }}auto{{ end }}'
     description: |
       Beszel is a lightweight server monitoring platform that includes Docker
       statistics, historical data, and alert functions. It has a friendly web

Makefile

@@ -0,0 +1,108 @@
+# Default OS/ARCH values
+OS ?= $(shell go env GOOS)
+ARCH ?= $(shell go env GOARCH)
+# Skip building the web UI if true
+SKIP_WEB ?= false
+
+# Set executable extension based on target OS
+EXE_EXT := $(if $(filter windows,$(OS)),.exe,)
+
+.PHONY: tidy build-agent build-hub build-hub-dev build clean lint dev-server dev-agent dev-hub dev generate-locales fetch-smartctl-conditional
+.DEFAULT_GOAL := build
+
+clean:
+	go clean
+	rm -rf ./build
+
+lint:
+	golangci-lint run
+
+test: export GOEXPERIMENT=synctest
+test:
+	go test -tags=testing ./...
+
+tidy:
+	go mod tidy
+
+build-web-ui:
+	@if command -v bun >/dev/null 2>&1; then \
+		bun install --cwd ./internal/site && \
+		bun run --cwd ./internal/site build; \
+	else \
+		npm install --prefix ./internal/site && \
+		npm run --prefix ./internal/site build; \
+	fi
+
+# Conditional .NET build - only for Windows
+build-dotnet-conditional:
+	@if [ "$(OS)" = "windows" ]; then \
+		echo "Building .NET executable for Windows..."; \
+		if command -v dotnet >/dev/null 2>&1; then \
+			rm -rf ./agent/lhm/bin; \
+			dotnet build -c Release ./agent/lhm/beszel_lhm.csproj; \
+		else \
+			echo "Error: dotnet not found. Install .NET SDK to build Windows agent."; \
+			exit 1; \
+		fi; \
+	fi
+
+# Download smartctl.exe at build time for Windows (skips if already present)
+fetch-smartctl-conditional:
+	@if [ "$(OS)" = "windows" ]; then \
+		go generate -run fetchsmartctl ./agent; \
+	fi
+
+# Update build-agent to include conditional .NET build
+build-agent: tidy build-dotnet-conditional fetch-smartctl-conditional
+	GOOS=$(OS) GOARCH=$(ARCH) go build -o ./build/beszel-agent_$(OS)_$(ARCH)$(EXE_EXT) -ldflags "-w -s" ./internal/cmd/agent
+
+build-hub: tidy $(if $(filter false,$(SKIP_WEB)),build-web-ui)
+	GOOS=$(OS) GOARCH=$(ARCH) go build -o ./build/beszel_$(OS)_$(ARCH)$(EXE_EXT) -ldflags "-w -s" ./internal/cmd/hub
+
+build-hub-dev: tidy
+	mkdir -p ./internal/site/dist && touch ./internal/site/dist/index.html
+	GOOS=$(OS) GOARCH=$(ARCH) go build -tags development -o ./build/beszel-dev_$(OS)_$(ARCH)$(EXE_EXT) -ldflags "-w -s" ./internal/cmd/hub
+
+build: build-agent build-hub
+
+generate-locales:
+	@if [ ! -f ./internal/site/src/locales/en/en.ts ]; then \
+		echo "Generating locales..."; \
+		command -v bun >/dev/null 2>&1 && cd ./internal/site && bun install && bun run sync || cd ./internal/site && npm install && npm run sync; \
+	fi
+
+dev-server: generate-locales
+	cd ./internal/site
+	@if command -v bun >/dev/null 2>&1; then \
+		cd ./internal/site && bun run dev --host 0.0.0.0; \
+	else \
+		cd ./internal/site && npm run dev --host 0.0.0.0; \
+	fi
+
+dev-hub: export ENV=dev
+dev-hub:
+	mkdir -p ./internal/site/dist && touch ./internal/site/dist/index.html
+	@if command -v entr >/dev/null 2>&1; then \
+		find ./internal -type f -name '*.go' | entr -r -s "cd ./internal/cmd/hub && go run -tags development . serve --http 0.0.0.0:8090"; \
+	else \
+		cd ./internal/cmd/hub && go run -tags development . serve --http 0.0.0.0:8090; \
+	fi
+
+dev-agent:
+	@if command -v entr >/dev/null 2>&1; then \
+		find ./internal/cmd/agent/*.go ./agent/*.go | entr -r go run github.com/henrygd/beszel/internal/cmd/agent; \
+	else \
+		go run github.com/henrygd/beszel/internal/cmd/agent; \
+	fi
+	
+build-dotnet:
+	@if command -v dotnet >/dev/null 2>&1; then \
+		rm -rf ./agent/lhm/bin; \
+		dotnet build -c Release ./agent/lhm/beszel_lhm.csproj; \
+	else \
+		echo "dotnet not found"; \
+	fi
+
+
+# KEY="..." make -j dev
+dev: dev-server dev-hub dev-agent

agent/agent.go

@@ -0,0 +1,265 @@
+// Package agent implements the Beszel monitoring agent that collects and serves system metrics.
+//
+// The agent runs on monitored systems and communicates collected data
+// to the Beszel hub for centralized monitoring and alerting.
+package agent
+
+import (
+	"crypto/sha256"
+	"encoding/hex"
+	"log/slog"
+	"os"
+	"path/filepath"
+	"strings"
+	"sync"
+	"time"
+
+	"github.com/gliderlabs/ssh"
+	"github.com/henrygd/beszel"
+	"github.com/henrygd/beszel/agent/deltatracker"
+	"github.com/henrygd/beszel/internal/common"
+	"github.com/henrygd/beszel/internal/entities/system"
+	"github.com/shirou/gopsutil/v4/host"
+	gossh "golang.org/x/crypto/ssh"
+)
+
+type Agent struct {
+	sync.Mutex                                                                      // Used to lock agent while collecting data
+	debug                     bool                                                  // true if LOG_LEVEL is set to debug
+	zfs                       bool                                                  // true if system has arcstats
+	memCalc                   string                                                // Memory calculation formula
+	fsNames                   []string                                              // List of filesystem device names being monitored
+	fsStats                   map[string]*system.FsStats                            // Keeps track of disk stats for each filesystem
+	diskPrev                  map[uint16]map[string]prevDisk                        // Previous disk I/O counters per cache interval
+	diskUsageCacheDuration    time.Duration                                         // How long to cache disk usage (to avoid waking sleeping disks)
+	lastDiskUsageUpdate       time.Time                                             // Last time disk usage was collected
+	netInterfaces             map[string]struct{}                                   // Stores all valid network interfaces
+	netIoStats                map[uint16]system.NetIoStats                          // Keeps track of bandwidth usage per cache interval
+	netInterfaceDeltaTrackers map[uint16]*deltatracker.DeltaTracker[string, uint64] // Per-cache-time NIC delta trackers
+	dockerManager             *dockerManager                                        // Manages Docker API requests
+	sensorConfig              *SensorConfig                                         // Sensors config
+	systemInfo                system.Info                                           // Host system info (dynamic)
+	systemDetails             system.Details                                        // Host system details (static, once-per-connection)
+	gpuManager                *GPUManager                                           // Manages GPU data
+	cache                     *systemDataCache                                      // Cache for system stats based on cache time
+	connectionManager         *ConnectionManager                                    // Channel to signal connection events
+	handlerRegistry           *HandlerRegistry                                      // Registry for routing incoming messages
+	server                    *ssh.Server                                           // SSH server
+	dataDir                   string                                                // Directory for persisting data
+	keys                      []gossh.PublicKey                                     // SSH public keys
+	smartManager              *SmartManager                                         // Manages SMART data
+	systemdManager            *systemdManager                                       // Manages systemd services
+}
+
+// NewAgent creates a new agent with the given data directory for persisting data.
+// If the data directory is not set, it will attempt to find the optimal directory.
+func NewAgent(dataDir ...string) (agent *Agent, err error) {
+	agent = &Agent{
+		fsStats: make(map[string]*system.FsStats),
+		cache:   NewSystemDataCache(),
+	}
+
+	// Initialize disk I/O previous counters storage
+	agent.diskPrev = make(map[uint16]map[string]prevDisk)
+	// Initialize per-cache-time network tracking structures
+	agent.netIoStats = make(map[uint16]system.NetIoStats)
+	agent.netInterfaceDeltaTrackers = make(map[uint16]*deltatracker.DeltaTracker[string, uint64])
+
+	agent.dataDir, err = getDataDir(dataDir...)
+	if err != nil {
+		slog.Warn("Data directory not found")
+	} else {
+		slog.Info("Data directory", "path", agent.dataDir)
+	}
+
+	agent.memCalc, _ = GetEnv("MEM_CALC")
+	agent.sensorConfig = agent.newSensorConfig()
+
+	// Parse disk usage cache duration (e.g., "15m", "1h") to avoid waking sleeping disks
+	if diskUsageCache, exists := GetEnv("DISK_USAGE_CACHE"); exists {
+		if duration, err := time.ParseDuration(diskUsageCache); err == nil {
+			agent.diskUsageCacheDuration = duration
+			slog.Info("DISK_USAGE_CACHE", "duration", duration)
+		} else {
+			slog.Warn("Invalid DISK_USAGE_CACHE", "err", err)
+		}
+	}
+
+	// Set up slog with a log level determined by the LOG_LEVEL env var
+	if logLevelStr, exists := GetEnv("LOG_LEVEL"); exists {
+		switch strings.ToLower(logLevelStr) {
+		case "debug":
+			agent.debug = true
+			slog.SetLogLoggerLevel(slog.LevelDebug)
+		case "warn":
+			slog.SetLogLoggerLevel(slog.LevelWarn)
+		case "error":
+			slog.SetLogLoggerLevel(slog.LevelError)
+		}
+	}
+
+	slog.Debug(beszel.Version)
+
+	// initialize docker manager
+	agent.dockerManager = newDockerManager()
+
+	// initialize system info
+	agent.refreshSystemDetails()
+
+	// SMART_INTERVAL env var to update smart data at this interval
+	if smartIntervalEnv, exists := GetEnv("SMART_INTERVAL"); exists {
+		if duration, err := time.ParseDuration(smartIntervalEnv); err == nil && duration > 0 {
+			agent.systemDetails.SmartInterval = duration
+			slog.Info("SMART_INTERVAL", "duration", duration)
+		} else {
+			slog.Warn("Invalid SMART_INTERVAL", "err", err)
+		}
+	}
+
+	// initialize connection manager
+	agent.connectionManager = newConnectionManager(agent)
+
+	// initialize handler registry
+	agent.handlerRegistry = NewHandlerRegistry()
+
+	// initialize disk info
+	agent.initializeDiskInfo()
+
+	// initialize net io stats
+	agent.initializeNetIoStats()
+
+	agent.systemdManager, err = newSystemdManager()
+	if err != nil {
+		slog.Debug("Systemd", "err", err)
+	}
+
+	agent.smartManager, err = NewSmartManager()
+	if err != nil {
+		slog.Debug("SMART", "err", err)
+	}
+
+	// initialize GPU manager
+	agent.gpuManager, err = NewGPUManager()
+	if err != nil {
+		slog.Debug("GPU", "err", err)
+	}
+
+	// if debugging, print stats
+	if agent.debug {
+		slog.Debug("Stats", "data", agent.gatherStats(common.DataRequestOptions{CacheTimeMs: 60_000, IncludeDetails: true}))
+	}
+
+	return agent, nil
+}
+
+// GetEnv retrieves an environment variable with a "BESZEL_AGENT_" prefix, or falls back to the unprefixed key.
+func GetEnv(key string) (value string, exists bool) {
+	if value, exists = os.LookupEnv("BESZEL_AGENT_" + key); exists {
+		return value, exists
+	}
+	// Fallback to the old unprefixed key
+	return os.LookupEnv(key)
+}
+
+func (a *Agent) gatherStats(options common.DataRequestOptions) *system.CombinedData {
+	a.Lock()
+	defer a.Unlock()
+
+	cacheTimeMs := options.CacheTimeMs
+	data, isCached := a.cache.Get(cacheTimeMs)
+	if isCached {
+		slog.Debug("Cached data", "cacheTimeMs", cacheTimeMs)
+		return data
+	}
+
+	*data = system.CombinedData{
+		Stats: a.getSystemStats(cacheTimeMs),
+		Info:  a.systemInfo,
+	}
+
+	// Include static system details only when requested
+	if options.IncludeDetails {
+		data.Details = &a.systemDetails
+	}
+
+	// slog.Info("System data", "data", data, "cacheTimeMs", cacheTimeMs)
+
+	if a.dockerManager != nil {
+		if containerStats, err := a.dockerManager.getDockerStats(cacheTimeMs); err == nil {
+			data.Containers = containerStats
+			slog.Debug("Containers", "data", data.Containers)
+		} else {
+			slog.Debug("Containers", "err", err)
+		}
+	}
+
+	// skip updating systemd services if cache time is not the default 60sec interval
+	if a.systemdManager != nil && cacheTimeMs == 60_000 {
+		totalCount := uint16(a.systemdManager.getServiceStatsCount())
+		if totalCount > 0 {
+			numFailed := a.systemdManager.getFailedServiceCount()
+			data.Info.Services = []uint16{totalCount, numFailed}
+		}
+		if a.systemdManager.hasFreshStats {
+			data.SystemdServices = a.systemdManager.getServiceStats(nil, false)
+		}
+	}
+
+	data.Stats.ExtraFs = make(map[string]*system.FsStats)
+	data.Info.ExtraFsPct = make(map[string]float64)
+	for name, stats := range a.fsStats {
+		if !stats.Root && stats.DiskTotal > 0 {
+			// Use custom name if available, otherwise use device name
+			key := name
+			if stats.Name != "" {
+				key = stats.Name
+			}
+			data.Stats.ExtraFs[key] = stats
+			// Add percentages to Info struct for dashboard
+			if stats.DiskTotal > 0 {
+				pct := twoDecimals((stats.DiskUsed / stats.DiskTotal) * 100)
+				data.Info.ExtraFsPct[key] = pct
+			}
+		}
+	}
+	slog.Debug("Extra FS", "data", data.Stats.ExtraFs)
+
+	a.cache.Set(data, cacheTimeMs)
+	return data
+}
+
+// StartAgent initializes and starts the agent with optional WebSocket connection
+func (a *Agent) Start(serverOptions ServerOptions) error {
+	a.keys = serverOptions.Keys
+	return a.connectionManager.Start(serverOptions)
+}
+
+func (a *Agent) getFingerprint() string {
+	// first look for a fingerprint in the data directory
+	if a.dataDir != "" {
+		if fp, err := os.ReadFile(filepath.Join(a.dataDir, "fingerprint")); err == nil {
+			return string(fp)
+		}
+	}
+
+	// if no fingerprint is found, generate one
+	fingerprint, err := host.HostID()
+	// we ignore a commonly known "product_uuid" known not to be unique
+	if err != nil || fingerprint == "" || fingerprint == "03000200-0400-0500-0006-000700080009" {
+		fingerprint = a.systemDetails.Hostname + a.systemDetails.CpuModel
+	}
+
+	// hash fingerprint
+	sum := sha256.Sum256([]byte(fingerprint))
+	fingerprint = hex.EncodeToString(sum[:24])
+
+	// save fingerprint to data directory
+	if a.dataDir != "" {
+		err = os.WriteFile(filepath.Join(a.dataDir, "fingerprint"), []byte(fingerprint), 0644)
+		if err != nil {
+			slog.Warn("Failed to save fingerprint", "err", err)
+		}
+	}
+
+	return fingerprint
+}

agent/agent_cache.go

@@ -0,0 +1,55 @@
+package agent
+
+import (
+	"sync"
+	"time"
+
+	"github.com/henrygd/beszel/internal/entities/system"
+)
+
+type systemDataCache struct {
+	sync.RWMutex
+	cache map[uint16]*cacheNode
+}
+
+type cacheNode struct {
+	data       *system.CombinedData
+	lastUpdate time.Time
+}
+
+// NewSystemDataCache creates a cache keyed by the polling interval in milliseconds.
+func NewSystemDataCache() *systemDataCache {
+	return &systemDataCache{
+		cache: make(map[uint16]*cacheNode),
+	}
+}
+
+// Get returns cached combined data when the entry is still considered fresh.
+func (c *systemDataCache) Get(cacheTimeMs uint16) (stats *system.CombinedData, isCached bool) {
+	c.RLock()
+	defer c.RUnlock()
+
+	node, ok := c.cache[cacheTimeMs]
+	if !ok {
+		return &system.CombinedData{}, false
+	}
+	// allowedSkew := time.Second
+	// isFresh := time.Since(node.lastUpdate) < time.Duration(cacheTimeMs)*time.Millisecond-allowedSkew
+	// allow a 50% skew of the cache time
+	isFresh := time.Since(node.lastUpdate) < time.Duration(cacheTimeMs/2)*time.Millisecond
+	return node.data, isFresh
+}
+
+// Set stores the latest combined data snapshot for the given interval.
+func (c *systemDataCache) Set(data *system.CombinedData, cacheTimeMs uint16) {
+	c.Lock()
+	defer c.Unlock()
+
+	node, ok := c.cache[cacheTimeMs]
+	if !ok {
+		node = &cacheNode{}
+		c.cache[cacheTimeMs] = node
+	}
+	node.data = data
+	node.lastUpdate = time.Now()
+}

agent/agent_cache_test.go

@@ -0,0 +1,246 @@
+//go:build testing
+// +build testing
+
+package agent
+
+import (
+	"testing"
+	"testing/synctest"
+	"time"
+
+	"github.com/henrygd/beszel/internal/entities/container"
+	"github.com/henrygd/beszel/internal/entities/system"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func createTestCacheData() *system.CombinedData {
+	return &system.CombinedData{
+		Stats: system.Stats{
+			Cpu:       50.5,
+			Mem:       8192,
+			DiskTotal: 100000,
+		},
+		Info: system.Info{
+			AgentVersion: "0.12.0",
+		},
+		Containers: []*container.Stats{
+			{
+				Name: "test-container",
+				Cpu:  25.0,
+			},
+		},
+	}
+}
+
+func TestNewSystemDataCache(t *testing.T) {
+	cache := NewSystemDataCache()
+	require.NotNil(t, cache)
+	assert.NotNil(t, cache.cache)
+	assert.Empty(t, cache.cache)
+}
+
+func TestCacheGetSet(t *testing.T) {
+	cache := NewSystemDataCache()
+	data := createTestCacheData()
+
+	// Test setting data
+	cache.Set(data, 1000) // 1 second cache
+
+	// Test getting fresh data
+	retrieved, isCached := cache.Get(1000)
+	assert.True(t, isCached)
+	assert.Equal(t, data, retrieved)
+
+	// Test getting non-existent cache key
+	_, isCached = cache.Get(2000)
+	assert.False(t, isCached)
+}
+
+func TestCacheFreshness(t *testing.T) {
+	cache := NewSystemDataCache()
+	data := createTestCacheData()
+
+	testCases := []struct {
+		name        string
+		cacheTimeMs uint16
+		sleepMs     time.Duration
+		expectFresh bool
+	}{
+		{
+			name:        "fresh data - well within cache time",
+			cacheTimeMs: 1000, // 1 second
+			sleepMs:     100,  // 100ms
+			expectFresh: true,
+		},
+		{
+			name:        "fresh data - at 50% of cache time boundary",
+			cacheTimeMs: 1000, // 1 second, 50% = 500ms
+			sleepMs:     499,  // just under 500ms
+			expectFresh: true,
+		},
+		{
+			name:        "stale data - exactly at 50% cache time",
+			cacheTimeMs: 1000, // 1 second, 50% = 500ms
+			sleepMs:     500,  // exactly 500ms
+			expectFresh: false,
+		},
+		{
+			name:        "stale data - well beyond cache time",
+			cacheTimeMs: 1000, // 1 second
+			sleepMs:     800,  // 800ms
+			expectFresh: false,
+		},
+		{
+			name:        "short cache time",
+			cacheTimeMs: 200, // 200ms, 50% = 100ms
+			sleepMs:     150, // 150ms > 100ms
+			expectFresh: false,
+		},
+	}
+
+	for _, tc := range testCases {
+		t.Run(tc.name, func(t *testing.T) {
+			synctest.Test(t, func(t *testing.T) {
+				// Set data
+				cache.Set(data, tc.cacheTimeMs)
+
+				// Wait for the specified duration
+				if tc.sleepMs > 0 {
+					time.Sleep(tc.sleepMs * time.Millisecond)
+				}
+
+				// Check freshness
+				_, isCached := cache.Get(tc.cacheTimeMs)
+				assert.Equal(t, tc.expectFresh, isCached)
+			})
+		})
+	}
+}
+
+func TestCacheMultipleIntervals(t *testing.T) {
+	synctest.Test(t, func(t *testing.T) {
+		cache := NewSystemDataCache()
+		data1 := createTestCacheData()
+		data2 := &system.CombinedData{
+			Stats: system.Stats{
+				Cpu: 75.0,
+				Mem: 16384,
+			},
+			Info: system.Info{
+				AgentVersion: "0.12.0",
+			},
+			Containers: []*container.Stats{},
+		}
+
+		// Set data for different intervals
+		cache.Set(data1, 500)  // 500ms cache
+		cache.Set(data2, 1000) // 1000ms cache
+
+		// Both should be fresh immediately
+		retrieved1, isCached1 := cache.Get(500)
+		assert.True(t, isCached1)
+		assert.Equal(t, data1, retrieved1)
+
+		retrieved2, isCached2 := cache.Get(1000)
+		assert.True(t, isCached2)
+		assert.Equal(t, data2, retrieved2)
+
+		// Wait 300ms - 500ms cache should be stale (250ms threshold), 1000ms should still be fresh (500ms threshold)
+		time.Sleep(300 * time.Millisecond)
+
+		_, isCached1 = cache.Get(500)
+		assert.False(t, isCached1)
+
+		_, isCached2 = cache.Get(1000)
+		assert.True(t, isCached2)
+
+		// Wait another 300ms (total 600ms) - now 1000ms cache should also be stale
+		time.Sleep(300 * time.Millisecond)
+		_, isCached2 = cache.Get(1000)
+		assert.False(t, isCached2)
+	})
+}
+
+func TestCacheOverwrite(t *testing.T) {
+	cache := NewSystemDataCache()
+	data1 := createTestCacheData()
+	data2 := &system.CombinedData{
+		Stats: system.Stats{
+			Cpu: 90.0,
+			Mem: 32768,
+		},
+		Info: system.Info{
+			AgentVersion: "0.12.0",
+		},
+		Containers: []*container.Stats{},
+	}
+
+	// Set initial data
+	cache.Set(data1, 1000)
+	retrieved, isCached := cache.Get(1000)
+	assert.True(t, isCached)
+	assert.Equal(t, data1, retrieved)
+
+	// Overwrite with new data
+	cache.Set(data2, 1000)
+	retrieved, isCached = cache.Get(1000)
+	assert.True(t, isCached)
+	assert.Equal(t, data2, retrieved)
+	assert.NotEqual(t, data1, retrieved)
+}
+
+func TestCacheMiss(t *testing.T) {
+	synctest.Test(t, func(t *testing.T) {
+		cache := NewSystemDataCache()
+
+		// Test getting from empty cache
+		_, isCached := cache.Get(1000)
+		assert.False(t, isCached)
+
+		// Set data for one interval
+		data := createTestCacheData()
+		cache.Set(data, 1000)
+
+		// Test getting different interval
+		_, isCached = cache.Get(2000)
+		assert.False(t, isCached)
+
+		// Test getting after data has expired
+		time.Sleep(600 * time.Millisecond) // 600ms > 500ms (50% of 1000ms)
+		_, isCached = cache.Get(1000)
+		assert.False(t, isCached)
+	})
+}
+
+func TestCacheZeroInterval(t *testing.T) {
+	cache := NewSystemDataCache()
+	data := createTestCacheData()
+
+	// Set with zero interval - should allow immediate cache
+	cache.Set(data, 0)
+
+	// With 0 interval, 50% is 0, so it should never be considered fresh
+	// (time.Since(lastUpdate) >= 0, which is not < 0)
+	_, isCached := cache.Get(0)
+	assert.False(t, isCached)
+}
+
+func TestCacheLargeInterval(t *testing.T) {
+	synctest.Test(t, func(t *testing.T) {
+		cache := NewSystemDataCache()
+		data := createTestCacheData()
+
+		// Test with maximum uint16 value
+		cache.Set(data, 65535) // ~65 seconds
+
+		// Should be fresh immediately
+		_, isCached := cache.Get(65535)
+		assert.True(t, isCached)
+
+		// Should still be fresh after a short time
+		time.Sleep(100 * time.Millisecond)
+		_, isCached = cache.Get(65535)
+		assert.True(t, isCached)
+	})
+}

agent/battery/battery.go

@@ -0,0 +1,84 @@
+//go:build !freebsd
+
+// Package battery provides functions to check if the system has a battery and to get the battery stats.
+package battery
+
+import (
+	"errors"
+	"log/slog"
+	"math"
+
+	"github.com/distatus/battery"
+)
+
+var (
+	systemHasBattery   = false
+	haveCheckedBattery = false
+)
+
+// HasReadableBattery checks if the system has a battery and returns true if it does.
+func HasReadableBattery() bool {
+	if haveCheckedBattery {
+		return systemHasBattery
+	}
+	haveCheckedBattery = true
+	batteries, err := battery.GetAll()
+	for _, bat := range batteries {
+		if bat != nil && (bat.Full > 0 || bat.Design > 0) {
+			systemHasBattery = true
+			break
+		}
+	}
+	if !systemHasBattery {
+		slog.Debug("No battery found", "err", err)
+	}
+	return systemHasBattery
+}
+
+// GetBatteryStats returns the current battery percent and charge state
+// percent = (current charge of all batteries) / (sum of designed/full capacity of all batteries)
+func GetBatteryStats() (batteryPercent uint8, batteryState uint8, err error) {
+	if !HasReadableBattery() {
+		return batteryPercent, batteryState, errors.ErrUnsupported
+	}
+	batteries, err := battery.GetAll()
+	// we'll handle errors later by skipping batteries with errors, rather
+	// than skipping everything because of the presence of some errors.
+	if len(batteries) == 0 {
+		return batteryPercent, batteryState, errors.New("no batteries")
+	}
+
+	totalCapacity := float64(0)
+	totalCharge := float64(0)
+	errs, partialErrs := err.(battery.Errors)
+
+	batteryState = math.MaxUint8
+
+	for i, bat := range batteries {
+		if partialErrs && errs[i] != nil {
+			// if there were some errors, like missing data, skip it
+			continue
+		}
+		if bat == nil || bat.Full == 0 {
+			// skip batteries with no capacity. Charge is unlikely to ever be zero, but
+			// we can't guarantee that, so don't skip based on charge.
+			continue
+		}
+		totalCapacity += bat.Full
+		totalCharge += bat.Current
+		if bat.State.Raw >= 0 {
+			batteryState = uint8(bat.State.Raw)
+		}
+	}
+
+	if totalCapacity == 0 || batteryState == math.MaxUint8 {
+		// for macs there's sometimes a ghost battery with 0 capacity
+		// https://github.com/distatus/battery/issues/34
+		// Instead of skipping over those batteries, we'll check for total 0 capacity
+		// and return an error. This also prevents a divide by zero.
+		return batteryPercent, batteryState, errors.New("no battery capacity")
+	}
+
+	batteryPercent = uint8(totalCharge / totalCapacity * 100)
+	return batteryPercent, batteryState, nil
+}

agent/client.go

@@ -1,8 +1,6 @@
 package agent
 
 import (
-	"beszel"
-	"beszel/internal/common"
 	"crypto/tls"
 	"errors"
 	"fmt"
@@ -15,6 +13,9 @@ import (
 	"strings"
 	"time"
 
+	"github.com/henrygd/beszel"
+	"github.com/henrygd/beszel/internal/common"
+
 	"github.com/fxamacker/cbor/v2"
 	"github.com/lxzan/gws"
 	"golang.org/x/crypto/ssh"
@@ -84,7 +85,7 @@ func getToken() (string, error) {
 	if err != nil {
 		return "", err
 	}
-	return string(tokenBytes), nil
+	return strings.TrimSpace(string(tokenBytes)), nil
 }
 
 // getOptions returns the WebSocket client options, creating them if necessary.
@@ -141,7 +142,9 @@ func (client *WebSocketClient) OnOpen(conn *gws.Conn) {
 // OnClose handles WebSocket connection closure.
 // It logs the closure reason and notifies the connection manager.
 func (client *WebSocketClient) OnClose(conn *gws.Conn, err error) {
-	slog.Warn("Connection closed", "err", strings.TrimPrefix(err.Error(), "gws: "))
+	if err != nil {
+		slog.Warn("Connection closed", "err", strings.TrimPrefix(err.Error(), "gws: "))
+	}
 	client.agent.connectionManager.eventChan <- WebSocketDisconnect
 }
 
@@ -155,11 +158,15 @@ func (client *WebSocketClient) OnMessage(conn *gws.Conn, message *gws.Message) {
 		return
 	}
 
-	if err := cbor.NewDecoder(message.Data).Decode(client.hubRequest); err != nil {
+	var HubRequest common.HubRequest[cbor.RawMessage]
+
+	err := cbor.Unmarshal(message.Data.Bytes(), &HubRequest)
+	if err != nil {
 		slog.Error("Error parsing message", "err", err)
 		return
 	}
-	if err := client.handleHubRequest(client.hubRequest); err != nil {
+
+	if err := client.handleHubRequest(&HubRequest, HubRequest.Id); err != nil {
 		slog.Error("Error handling message", "err", err)
 	}
 }
@@ -172,7 +179,7 @@ func (client *WebSocketClient) OnPing(conn *gws.Conn, message []byte) {
 }
 
 // handleAuthChallenge verifies the authenticity of the hub and returns the system's fingerprint.
-func (client *WebSocketClient) handleAuthChallenge(msg *common.HubRequest[cbor.RawMessage]) (err error) {
+func (client *WebSocketClient) handleAuthChallenge(msg *common.HubRequest[cbor.RawMessage], requestID *uint32) (err error) {
 	var authRequest common.FingerprintRequest
 	if err := cbor.Unmarshal(msg.Data, &authRequest); err != nil {
 		return err
@@ -190,12 +197,13 @@ func (client *WebSocketClient) handleAuthChallenge(msg *common.HubRequest[cbor.R
 	}
 
 	if authRequest.NeedSysInfo {
-		response.Hostname = client.agent.systemInfo.Hostname
+		response.Name, _ = GetEnv("SYSTEM_NAME")
+		response.Hostname = client.agent.systemDetails.Hostname
 		serverAddr := client.agent.connectionManager.serverOptions.Addr
 		_, response.Port, _ = net.SplitHostPort(serverAddr)
 	}
 
-	return client.sendMessage(response)
+	return client.sendResponse(response, requestID)
 }
 
 // verifySignature verifies the signature of the token using the public keys.
@@ -220,25 +228,17 @@ func (client *WebSocketClient) Close() {
 	}
 }
 
-// handleHubRequest routes the request to the appropriate handler.
-// It ensures the hub is verified before processing most requests.
-func (client *WebSocketClient) handleHubRequest(msg *common.HubRequest[cbor.RawMessage]) error {
-	if !client.hubVerified && msg.Action != common.CheckFingerprint {
-		return errors.New("hub not verified")
+// handleHubRequest routes the request to the appropriate handler using the handler registry.
+func (client *WebSocketClient) handleHubRequest(msg *common.HubRequest[cbor.RawMessage], requestID *uint32) error {
+	ctx := &HandlerContext{
+		Client:       client,
+		Agent:        client.agent,
+		Request:      msg,
+		RequestID:    requestID,
+		HubVerified:  client.hubVerified,
+		SendResponse: client.sendResponse,
 	}
-	switch msg.Action {
-	case common.GetData:
-		return client.sendSystemData()
-	case common.CheckFingerprint:
-		return client.handleAuthChallenge(msg)
-	}
-	return nil
-}
-
-// sendSystemData gathers and sends current system statistics to the hub.
-func (client *WebSocketClient) sendSystemData() error {
-	sysStats := client.agent.gatherStats(client.token)
-	return client.sendMessage(sysStats)
+	return client.agent.handlerRegistry.Handle(ctx)
 }
 
 // sendMessage encodes the given data to CBOR and sends it as a binary message over the WebSocket connection to the hub.
@@ -247,7 +247,25 @@ func (client *WebSocketClient) sendMessage(data any) error {
 	if err != nil {
 		return err
 	}
-	return client.Conn.WriteMessage(gws.OpcodeBinary, bytes)
+	err = client.Conn.WriteMessage(gws.OpcodeBinary, bytes)
+	if err != nil {
+		// If writing fails (e.g., broken pipe due to network issues),
+		// close the connection to trigger reconnection logic (#1263)
+		client.Close()
+	}
+	return err
+}
+
+// sendResponse sends a response with optional request ID.
+// For ID-based requests, we must populate legacy typed fields for backward
+// compatibility with older hubs (<= 0.17) that don't read the generic Data field.
+func (client *WebSocketClient) sendResponse(data any, requestID *uint32) error {
+	if requestID != nil {
+		response := newAgentResponse(data, requestID)
+		return client.sendMessage(response)
+	}
+	// Legacy format - send data directly
+	return client.sendMessage(data)
 }
 
 // getUserAgent returns one of two User-Agent strings based on current time.

agent/client_test.go

@@ -4,8 +4,6 @@
 package agent
 
 import (
-	"beszel"
-	"beszel/internal/common"
 	"crypto/ed25519"
 	"net/url"
 	"os"
@@ -13,6 +11,10 @@ import (
 	"testing"
 	"time"
 
+	"github.com/henrygd/beszel"
+
+	"github.com/henrygd/beszel/internal/common"
+
 	"github.com/fxamacker/cbor/v2"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
@@ -299,7 +301,7 @@ func TestWebSocketClient_HandleHubRequest(t *testing.T) {
 				Data:   cbor.RawMessage{},
 			}
 
-			err := client.handleHubRequest(hubRequest)
+			err := client.handleHubRequest(hubRequest, nil)
 
 			if tc.expectError {
 				assert.Error(t, err)
@@ -535,4 +537,25 @@ func TestGetToken(t *testing.T) {
 		assert.NoError(t, err)
 		assert.Equal(t, "", token, "Empty file should return empty string")
 	})
+
+	t.Run("strips whitespace from TOKEN_FILE", func(t *testing.T) {
+		unsetEnvVars()
+
+		tokenWithWhitespace := "  test-token-with-whitespace  \n\t"
+		expectedToken := "test-token-with-whitespace"
+		tokenFile, err := os.CreateTemp("", "token-test-*.txt")
+		require.NoError(t, err)
+		defer os.Remove(tokenFile.Name())
+
+		_, err = tokenFile.WriteString(tokenWithWhitespace)
+		require.NoError(t, err)
+		tokenFile.Close()
+
+		os.Setenv("TOKEN_FILE", tokenFile.Name())
+		defer os.Unsetenv("TOKEN_FILE")
+
+		token, err := getToken()
+		assert.NoError(t, err)
+		assert.Equal(t, expectedToken, token, "Whitespace should be stripped from token file content")
+	})
 }

agent/connection_manager.go

@@ -1,26 +1,29 @@
 package agent
 
 import (
-	"beszel/internal/agent/health"
 	"errors"
 	"log/slog"
 	"os"
 	"os/signal"
 	"syscall"
 	"time"
+
+	"github.com/henrygd/beszel/agent/health"
+	"github.com/henrygd/beszel/internal/entities/system"
 )
 
 // ConnectionManager manages the connection state and events for the agent.
 // It handles both WebSocket and SSH connections, automatically switching between
 // them based on availability and managing reconnection attempts.
 type ConnectionManager struct {
-	agent         *Agent               // Reference to the parent agent
-	State         ConnectionState      // Current connection state
-	eventChan     chan ConnectionEvent // Channel for connection events
-	wsClient      *WebSocketClient     // WebSocket client for hub communication
-	serverOptions ServerOptions        // Configuration for SSH server
-	wsTicker      *time.Ticker         // Ticker for WebSocket connection attempts
-	isConnecting  bool                 // Prevents multiple simultaneous reconnection attempts
+	agent          *Agent               // Reference to the parent agent
+	State          ConnectionState      // Current connection state
+	eventChan      chan ConnectionEvent // Channel for connection events
+	wsClient       *WebSocketClient     // WebSocket client for hub communication
+	serverOptions  ServerOptions        // Configuration for SSH server
+	wsTicker       *time.Ticker         // Ticker for WebSocket connection attempts
+	isConnecting   bool                 // Prevents multiple simultaneous reconnection attempts
+	ConnectionType system.ConnectionType
 }
 
 // ConnectionState represents the current connection state of the agent.
@@ -143,15 +146,18 @@ func (c *ConnectionManager) handleStateChange(newState ConnectionState) {
 	switch newState {
 	case WebSocketConnected:
 		slog.Info("WebSocket connected", "host", c.wsClient.hubURL.Host)
+		c.ConnectionType = system.ConnectionTypeWebSocket
 		c.stopWsTicker()
 		_ = c.agent.StopServer()
 		c.isConnecting = false
 	case SSHConnected:
 		// stop new ws connection attempts
 		slog.Info("SSH connection established")
+		c.ConnectionType = system.ConnectionTypeSSH
 		c.stopWsTicker()
 		c.isConnecting = false
 	case Disconnected:
+		c.ConnectionType = system.ConnectionTypeNone
 		if c.isConnecting {
 			// Already handling reconnection, avoid duplicate attempts
 			return

agent/cpu.go

@@ -0,0 +1,134 @@
+package agent
+
+import (
+	"math"
+	"runtime"
+
+	"github.com/henrygd/beszel/internal/entities/system"
+	"github.com/shirou/gopsutil/v4/cpu"
+)
+
+var lastCpuTimes = make(map[uint16]cpu.TimesStat)
+var lastPerCoreCpuTimes = make(map[uint16][]cpu.TimesStat)
+
+// init initializes the CPU monitoring by storing the initial CPU times
+// for the default 60-second cache interval.
+func init() {
+	if times, err := cpu.Times(false); err == nil {
+		lastCpuTimes[60000] = times[0]
+	}
+	if perCoreTimes, err := cpu.Times(true); err == nil {
+		lastPerCoreCpuTimes[60000] = perCoreTimes
+	}
+}
+
+// CpuMetrics contains detailed CPU usage breakdown
+type CpuMetrics struct {
+	Total  float64
+	User   float64
+	System float64
+	Iowait float64
+	Steal  float64
+	Idle   float64
+}
+
+// getCpuMetrics calculates detailed CPU usage metrics using cached previous measurements.
+// It returns percentages for total, user, system, iowait, and steal time.
+func getCpuMetrics(cacheTimeMs uint16) (CpuMetrics, error) {
+	times, err := cpu.Times(false)
+	if err != nil || len(times) == 0 {
+		return CpuMetrics{}, err
+	}
+	// if cacheTimeMs is not in lastCpuTimes, use 60000 as fallback lastCpuTime
+	if _, ok := lastCpuTimes[cacheTimeMs]; !ok {
+		lastCpuTimes[cacheTimeMs] = lastCpuTimes[60000]
+	}
+
+	t1 := lastCpuTimes[cacheTimeMs]
+	t2 := times[0]
+
+	t1All, _ := getAllBusy(t1)
+	t2All, _ := getAllBusy(t2)
+
+	totalDelta := t2All - t1All
+	if totalDelta <= 0 {
+		return CpuMetrics{}, nil
+	}
+
+	metrics := CpuMetrics{
+		Total:  calculateBusy(t1, t2),
+		User:   clampPercent((t2.User - t1.User) / totalDelta * 100),
+		System: clampPercent((t2.System - t1.System) / totalDelta * 100),
+		Iowait: clampPercent((t2.Iowait - t1.Iowait) / totalDelta * 100),
+		Steal:  clampPercent((t2.Steal - t1.Steal) / totalDelta * 100),
+		Idle:   clampPercent((t2.Idle - t1.Idle) / totalDelta * 100),
+	}
+
+	lastCpuTimes[cacheTimeMs] = times[0]
+	return metrics, nil
+}
+
+// clampPercent ensures the percentage is between 0 and 100
+func clampPercent(value float64) float64 {
+	return math.Min(100, math.Max(0, value))
+}
+
+// getPerCoreCpuUsage calculates per-core CPU busy usage as integer percentages (0-100).
+// It uses cached previous measurements for the provided cache interval.
+func getPerCoreCpuUsage(cacheTimeMs uint16) (system.Uint8Slice, error) {
+	perCoreTimes, err := cpu.Times(true)
+	if err != nil || len(perCoreTimes) == 0 {
+		return nil, err
+	}
+
+	// Initialize cache if needed
+	if _, ok := lastPerCoreCpuTimes[cacheTimeMs]; !ok {
+		lastPerCoreCpuTimes[cacheTimeMs] = lastPerCoreCpuTimes[60000]
+	}
+
+	lastTimes := lastPerCoreCpuTimes[cacheTimeMs]
+
+	// Limit to the number of cores available in both samples
+	length := len(perCoreTimes)
+	if len(lastTimes) < length {
+		length = len(lastTimes)
+	}
+
+	usage := make([]uint8, length)
+	for i := 0; i < length; i++ {
+		t1 := lastTimes[i]
+		t2 := perCoreTimes[i]
+		usage[i] = uint8(math.Round(calculateBusy(t1, t2)))
+	}
+
+	lastPerCoreCpuTimes[cacheTimeMs] = perCoreTimes
+	return usage, nil
+}
+
+// calculateBusy calculates the CPU busy percentage between two time points.
+// It computes the ratio of busy time to total time elapsed between t1 and t2,
+// returning a percentage clamped between 0 and 100.
+func calculateBusy(t1, t2 cpu.TimesStat) float64 {
+	t1All, t1Busy := getAllBusy(t1)
+	t2All, t2Busy := getAllBusy(t2)
+
+	if t2All <= t1All || t2Busy <= t1Busy {
+		return 0
+	}
+	return clampPercent((t2Busy - t1Busy) / (t2All - t1All) * 100)
+}
+
+// getAllBusy calculates the total CPU time and busy CPU time from CPU times statistics.
+// On Linux, it excludes guest and guest_nice time from the total to match kernel behavior.
+// Returns total CPU time and busy CPU time (total minus idle and I/O wait time).
+func getAllBusy(t cpu.TimesStat) (float64, float64) {
+	tot := t.Total()
+	if runtime.GOOS == "linux" {
+		tot -= t.Guest     // Linux 2.6.24+
+		tot -= t.GuestNice // Linux 3.2.0+
+	}
+
+	busy := tot - t.Idle - t.Iowait
+
+	return tot, busy
+}

agent/deltatracker/deltatracker.go

@@ -0,0 +1,100 @@
+// Package deltatracker provides a tracker for calculating differences in numeric values over time.
+package deltatracker
+
+import (
+	"sync"
+
+	"golang.org/x/exp/constraints"
+)
+
+// Numeric is a constraint that permits any integer or floating-point type.
+type Numeric interface {
+	constraints.Integer | constraints.Float
+}
+
+// DeltaTracker is a generic, thread-safe tracker for calculating differences
+// in numeric values over time.
+// K is the key type (e.g., int, string).
+// V is the value type (e.g., int, int64, float32, float64).
+type DeltaTracker[K comparable, V Numeric] struct {
+	sync.RWMutex
+	current  map[K]V
+	previous map[K]V
+}
+
+// NewDeltaTracker creates a new generic tracker.
+func NewDeltaTracker[K comparable, V Numeric]() *DeltaTracker[K, V] {
+	return &DeltaTracker[K, V]{
+		current:  make(map[K]V),
+		previous: make(map[K]V),
+	}
+}
+
+// Set records the current value for a given ID.
+func (t *DeltaTracker[K, V]) Set(id K, value V) {
+	t.Lock()
+	defer t.Unlock()
+	t.current[id] = value
+}
+
+// Snapshot returns a copy of the current map.
+// func (t *DeltaTracker[K, V]) Snapshot() map[K]V {
+// 	t.RLock()
+// 	defer t.RUnlock()
+
+// 	copyMap := make(map[K]V, len(t.current))
+// 	maps.Copy(copyMap, t.current)
+// 	return copyMap
+// }
+
+// Deltas returns a map of all calculated deltas for the current interval.
+func (t *DeltaTracker[K, V]) Deltas() map[K]V {
+	t.RLock()
+	defer t.RUnlock()
+
+	deltas := make(map[K]V)
+	for id, currentVal := range t.current {
+		if previousVal, ok := t.previous[id]; ok {
+			deltas[id] = currentVal - previousVal
+		} else {
+			deltas[id] = 0
+		}
+	}
+	return deltas
+}
+
+// Previous returns the previously recorded value for the given key, if it exists.
+func (t *DeltaTracker[K, V]) Previous(id K) (V, bool) {
+	t.RLock()
+	defer t.RUnlock()
+
+	value, ok := t.previous[id]
+	return value, ok
+}
+
+// Delta returns the delta for a single key.
+// Returns 0 if the key doesn't exist or has no previous value.
+func (t *DeltaTracker[K, V]) Delta(id K) V {
+	t.RLock()
+	defer t.RUnlock()
+
+	currentVal, currentOk := t.current[id]
+	if !currentOk {
+		return 0
+	}
+
+	previousVal, previousOk := t.previous[id]
+	if !previousOk {
+		return 0
+	}
+
+	return currentVal - previousVal
+}
+
+// Cycle prepares the tracker for the next interval.
+func (t *DeltaTracker[K, V]) Cycle() {
+	t.Lock()
+	defer t.Unlock()
+	t.previous = t.current
+	t.current = make(map[K]V)
+}

agent/deltatracker/deltatracker_test.go

@@ -0,0 +1,217 @@
+package deltatracker
+
+import (
+	"fmt"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func ExampleDeltaTracker() {
+	tracker := NewDeltaTracker[string, int]()
+	tracker.Set("key1", 10)
+	tracker.Set("key2", 20)
+	tracker.Cycle()
+	tracker.Set("key1", 15)
+	tracker.Set("key2", 30)
+	fmt.Println(tracker.Delta("key1"))
+	fmt.Println(tracker.Delta("key2"))
+	fmt.Println(tracker.Deltas())
+	// Output: 5
+	// 10
+	// map[key1:5 key2:10]
+}
+
+func TestNewDeltaTracker(t *testing.T) {
+	tracker := NewDeltaTracker[string, int]()
+	assert.NotNil(t, tracker)
+	assert.Empty(t, tracker.current)
+	assert.Empty(t, tracker.previous)
+}
+
+func TestSet(t *testing.T) {
+	tracker := NewDeltaTracker[string, int]()
+	tracker.Set("key1", 10)
+
+	tracker.RLock()
+	defer tracker.RUnlock()
+
+	assert.Equal(t, 10, tracker.current["key1"])
+}
+
+func TestDeltas(t *testing.T) {
+	tracker := NewDeltaTracker[string, int]()
+
+	// Test with no previous values
+	tracker.Set("key1", 10)
+	tracker.Set("key2", 20)
+
+	deltas := tracker.Deltas()
+	assert.Equal(t, 0, deltas["key1"])
+	assert.Equal(t, 0, deltas["key2"])
+
+	// Cycle to move current to previous
+	tracker.Cycle()
+
+	// Set new values and check deltas
+	tracker.Set("key1", 15) // Delta should be 5 (15-10)
+	tracker.Set("key2", 25) // Delta should be 5 (25-20)
+	tracker.Set("key3", 30) // New key, delta should be 0
+
+	deltas = tracker.Deltas()
+	assert.Equal(t, 5, deltas["key1"])
+	assert.Equal(t, 5, deltas["key2"])
+	assert.Equal(t, 0, deltas["key3"])
+}
+
+func TestCycle(t *testing.T) {
+	tracker := NewDeltaTracker[string, int]()
+
+	tracker.Set("key1", 10)
+	tracker.Set("key2", 20)
+
+	// Verify current has values
+	tracker.RLock()
+	assert.Equal(t, 10, tracker.current["key1"])
+	assert.Equal(t, 20, tracker.current["key2"])
+	assert.Empty(t, tracker.previous)
+	tracker.RUnlock()
+
+	tracker.Cycle()
+
+	// After cycle, previous should have the old current values
+	// and current should be empty
+	tracker.RLock()
+	assert.Empty(t, tracker.current)
+	assert.Equal(t, 10, tracker.previous["key1"])
+	assert.Equal(t, 20, tracker.previous["key2"])
+	tracker.RUnlock()
+}
+
+func TestCompleteWorkflow(t *testing.T) {
+	tracker := NewDeltaTracker[string, int]()
+
+	// First interval
+	tracker.Set("server1", 100)
+	tracker.Set("server2", 200)
+
+	// Get deltas for first interval (should be zero)
+	firstDeltas := tracker.Deltas()
+	assert.Equal(t, 0, firstDeltas["server1"])
+	assert.Equal(t, 0, firstDeltas["server2"])
+
+	// Cycle to next interval
+	tracker.Cycle()
+
+	// Second interval
+	tracker.Set("server1", 150) // Delta: 50
+	tracker.Set("server2", 180) // Delta: -20
+	tracker.Set("server3", 300) // New server, delta: 300
+
+	secondDeltas := tracker.Deltas()
+	assert.Equal(t, 50, secondDeltas["server1"])
+	assert.Equal(t, -20, secondDeltas["server2"])
+	assert.Equal(t, 0, secondDeltas["server3"])
+}
+
+func TestDeltaTrackerWithDifferentTypes(t *testing.T) {
+	// Test with int64
+	intTracker := NewDeltaTracker[string, int64]()
+	intTracker.Set("pid1", 1000)
+	intTracker.Cycle()
+	intTracker.Set("pid1", 1200)
+	intDeltas := intTracker.Deltas()
+	assert.Equal(t, int64(200), intDeltas["pid1"])
+
+	// Test with float64
+	floatTracker := NewDeltaTracker[string, float64]()
+	floatTracker.Set("cpu1", 1.5)
+	floatTracker.Cycle()
+	floatTracker.Set("cpu1", 2.7)
+	floatDeltas := floatTracker.Deltas()
+	assert.InDelta(t, 1.2, floatDeltas["cpu1"], 0.0001)
+
+	// Test with int keys
+	pidTracker := NewDeltaTracker[int, int64]()
+	pidTracker.Set(101, 20000)
+	pidTracker.Cycle()
+	pidTracker.Set(101, 22500)
+	pidDeltas := pidTracker.Deltas()
+	assert.Equal(t, int64(2500), pidDeltas[101])
+}
+
+func TestDelta(t *testing.T) {
+	tracker := NewDeltaTracker[string, int]()
+
+	// Test getting delta for non-existent key
+	result := tracker.Delta("nonexistent")
+	assert.Equal(t, 0, result)
+
+	// Test getting delta for key with no previous value
+	tracker.Set("key1", 10)
+	result = tracker.Delta("key1")
+	assert.Equal(t, 0, result)
+
+	// Cycle to move current to previous
+	tracker.Cycle()
+
+	// Test getting delta for key with previous value
+	tracker.Set("key1", 15)
+	result = tracker.Delta("key1")
+	assert.Equal(t, 5, result)
+
+	// Test getting delta for key that exists in previous but not current
+	result = tracker.Delta("key1")
+	assert.Equal(t, 5, result) // Should still return 5
+
+	// Test getting delta for key that exists in current but not previous
+	tracker.Set("key2", 20)
+	result = tracker.Delta("key2")
+	assert.Equal(t, 0, result)
+}
+
+func TestDeltaWithDifferentTypes(t *testing.T) {
+	// Test with int64
+	intTracker := NewDeltaTracker[string, int64]()
+	intTracker.Set("pid1", 1000)
+	intTracker.Cycle()
+	intTracker.Set("pid1", 1200)
+	result := intTracker.Delta("pid1")
+	assert.Equal(t, int64(200), result)
+
+	// Test with float64
+	floatTracker := NewDeltaTracker[string, float64]()
+	floatTracker.Set("cpu1", 1.5)
+	floatTracker.Cycle()
+	floatTracker.Set("cpu1", 2.7)
+	floatResult := floatTracker.Delta("cpu1")
+	assert.InDelta(t, 1.2, floatResult, 0.0001)
+
+	// Test with int keys
+	pidTracker := NewDeltaTracker[int, int64]()
+	pidTracker.Set(101, 20000)
+	pidTracker.Cycle()
+	pidTracker.Set(101, 22500)
+	pidResult := pidTracker.Delta(101)
+	assert.Equal(t, int64(2500), pidResult)
+}
+
+func TestDeltaConcurrentAccess(t *testing.T) {
+	tracker := NewDeltaTracker[string, int]()
+
+	// Set initial values
+	tracker.Set("key1", 10)
+	tracker.Set("key2", 20)
+	tracker.Cycle()
+
+	// Set new values
+	tracker.Set("key1", 15)
+	tracker.Set("key2", 25)
+
+	// Test concurrent access safety
+	result1 := tracker.Delta("key1")
+	result2 := tracker.Delta("key2")
+
+	assert.Equal(t, 5, result1)
+	assert.Equal(t, 5, result2)
+}

agent/disk.go

@@ -0,0 +1,362 @@
+package agent
+
+import (
+	"log/slog"
+	"os"
+	"path/filepath"
+	"runtime"
+	"strings"
+	"time"
+
+	"github.com/henrygd/beszel/internal/entities/system"
+
+	"github.com/shirou/gopsutil/v4/disk"
+)
+
+// parseFilesystemEntry parses a filesystem entry in the format "device__customname"
+// Returns the device/filesystem part and the custom name part
+func parseFilesystemEntry(entry string) (device, customName string) {
+	entry = strings.TrimSpace(entry)
+	if parts := strings.SplitN(entry, "__", 2); len(parts) == 2 {
+		device = strings.TrimSpace(parts[0])
+		customName = strings.TrimSpace(parts[1])
+	} else {
+		device = entry
+	}
+	return device, customName
+}
+
+// Sets up the filesystems to monitor for disk usage and I/O.
+func (a *Agent) initializeDiskInfo() {
+	filesystem, _ := GetEnv("FILESYSTEM")
+	efPath := "/extra-filesystems"
+	hasRoot := false
+	isWindows := runtime.GOOS == "windows"
+
+	partitions, err := disk.Partitions(false)
+	if err != nil {
+		slog.Error("Error getting disk partitions", "err", err)
+	}
+	slog.Debug("Disk", "partitions", partitions)
+
+	// trim trailing backslash for Windows devices (#1361)
+	if isWindows {
+		for i, p := range partitions {
+			partitions[i].Device = strings.TrimSuffix(p.Device, "\\")
+		}
+	}
+
+	// ioContext := context.WithValue(a.sensorsContext,
+	// 	common.EnvKey, common.EnvMap{common.HostProcEnvKey: "/tmp/testproc"},
+	// )
+	// diskIoCounters, err := disk.IOCountersWithContext(ioContext)
+
+	diskIoCounters, err := disk.IOCounters()
+	if err != nil {
+		slog.Error("Error getting diskstats", "err", err)
+	}
+	slog.Debug("Disk I/O", "diskstats", diskIoCounters)
+
+	// Helper function to add a filesystem to fsStats if it doesn't exist
+	addFsStat := func(device, mountpoint string, root bool, customName ...string) {
+		var key string
+		if isWindows {
+			key = device
+		} else {
+			key = filepath.Base(device)
+		}
+		var ioMatch bool
+		if _, exists := a.fsStats[key]; !exists {
+			if root {
+				slog.Info("Detected root device", "name", key)
+				// Check if root device is in /proc/diskstats, use fallback if not
+				if _, ioMatch = diskIoCounters[key]; !ioMatch {
+					key, ioMatch = findIoDevice(filesystem, diskIoCounters, a.fsStats)
+					if !ioMatch {
+						slog.Info("Using I/O fallback", "device", device, "mountpoint", mountpoint, "fallback", key)
+					}
+				}
+			} else {
+				// Check if non-root has diskstats and fall back to folder name if not
+				// Scenario: device is encrypted and named luks-2bcb02be-999d-4417-8d18-5c61e660fb6e - not in /proc/diskstats.
+				// However, the device can be specified by mounting folder from luks device at /extra-filesystems/sda1
+				if _, ioMatch = diskIoCounters[key]; !ioMatch {
+					efBase := filepath.Base(mountpoint)
+					if _, ioMatch = diskIoCounters[efBase]; ioMatch {
+						key = efBase
+					}
+				}
+			}
+			fsStats := &system.FsStats{Root: root, Mountpoint: mountpoint}
+			if len(customName) > 0 && customName[0] != "" {
+				fsStats.Name = customName[0]
+			}
+			a.fsStats[key] = fsStats
+		}
+	}
+
+	// Get the appropriate root mount point for this system
+	rootMountPoint := a.getRootMountPoint()
+
+	// Use FILESYSTEM env var to find root filesystem
+	if filesystem != "" {
+		for _, p := range partitions {
+			if strings.HasSuffix(p.Device, filesystem) || p.Mountpoint == filesystem {
+				addFsStat(p.Device, p.Mountpoint, true)
+				hasRoot = true
+				break
+			}
+		}
+		if !hasRoot {
+			slog.Warn("Partition details not found", "filesystem", filesystem)
+		}
+	}
+
+	// Add EXTRA_FILESYSTEMS env var values to fsStats
+	if extraFilesystems, exists := GetEnv("EXTRA_FILESYSTEMS"); exists {
+		for _, fsEntry := range strings.Split(extraFilesystems, ",") {
+			// Parse custom name from format: device__customname
+			fs, customName := parseFilesystemEntry(fsEntry)
+
+			found := false
+			for _, p := range partitions {
+				if strings.HasSuffix(p.Device, fs) || p.Mountpoint == fs {
+					addFsStat(p.Device, p.Mountpoint, false, customName)
+					found = true
+					break
+				}
+			}
+			// if not in partitions, test if we can get disk usage
+			if !found {
+				if _, err := disk.Usage(fs); err == nil {
+					addFsStat(filepath.Base(fs), fs, false, customName)
+				} else {
+					slog.Error("Invalid filesystem", "name", fs, "err", err)
+				}
+			}
+		}
+	}
+
+	// Process partitions for various mount points
+	for _, p := range partitions {
+		// fmt.Println(p.Device, p.Mountpoint)
+		// Binary root fallback or docker root fallback
+		if !hasRoot && (p.Mountpoint == rootMountPoint || (p.Mountpoint == "/etc/hosts" && strings.HasPrefix(p.Device, "/dev"))) {
+			fs, match := findIoDevice(filepath.Base(p.Device), diskIoCounters, a.fsStats)
+			if match {
+				addFsStat(fs, p.Mountpoint, true)
+				hasRoot = true
+			}
+		}
+
+		// Check if device is in /extra-filesystems
+		if strings.HasPrefix(p.Mountpoint, efPath) {
+			device, customName := parseFilesystemEntry(p.Mountpoint)
+			addFsStat(device, p.Mountpoint, false, customName)
+		}
+	}
+
+	// Check all folders in /extra-filesystems and add them if not already present
+	if folders, err := os.ReadDir(efPath); err == nil {
+		existingMountpoints := make(map[string]bool)
+		for _, stats := range a.fsStats {
+			existingMountpoints[stats.Mountpoint] = true
+		}
+		for _, folder := range folders {
+			if folder.IsDir() {
+				mountpoint := filepath.Join(efPath, folder.Name())
+				slog.Debug("/extra-filesystems", "mountpoint", mountpoint)
+				if !existingMountpoints[mountpoint] {
+					device, customName := parseFilesystemEntry(folder.Name())
+					addFsStat(device, mountpoint, false, customName)
+				}
+			}
+		}
+	}
+
+	// If no root filesystem set, use fallback
+	if !hasRoot {
+		rootDevice, _ := findIoDevice(filepath.Base(filesystem), diskIoCounters, a.fsStats)
+		slog.Info("Root disk", "mountpoint", rootMountPoint, "io", rootDevice)
+		a.fsStats[rootDevice] = &system.FsStats{Root: true, Mountpoint: rootMountPoint}
+	}
+
+	a.initializeDiskIoStats(diskIoCounters)
+}
+
+// Returns matching device from /proc/diskstats,
+// or the device with the most reads if no match is found.
+// bool is true if a match was found.
+func findIoDevice(filesystem string, diskIoCounters map[string]disk.IOCountersStat, fsStats map[string]*system.FsStats) (string, bool) {
+	var maxReadBytes uint64
+	maxReadDevice := "/"
+	for _, d := range diskIoCounters {
+		if d.Name == filesystem || (d.Label != "" && d.Label == filesystem) {
+			return d.Name, true
+		}
+		if d.ReadBytes > maxReadBytes {
+			// don't use if device already exists in fsStats
+			if _, exists := fsStats[d.Name]; !exists {
+				maxReadBytes = d.ReadBytes
+				maxReadDevice = d.Name
+			}
+		}
+	}
+	return maxReadDevice, false
+}
+
+// Sets start values for disk I/O stats.
+func (a *Agent) initializeDiskIoStats(diskIoCounters map[string]disk.IOCountersStat) {
+	for device, stats := range a.fsStats {
+		// skip if not in diskIoCounters
+		d, exists := diskIoCounters[device]
+		if !exists {
+			slog.Warn("Device not found in diskstats", "name", device)
+			continue
+		}
+		// populate initial values
+		stats.Time = time.Now()
+		stats.TotalRead = d.ReadBytes
+		stats.TotalWrite = d.WriteBytes
+		// add to list of valid io device names
+		a.fsNames = append(a.fsNames, device)
+	}
+}
+
+// Updates disk usage statistics for all monitored filesystems
+func (a *Agent) updateDiskUsage(systemStats *system.Stats) {
+	// Check if we should skip extra filesystem collection to avoid waking sleeping disks.
+	// Root filesystem is always updated since it can't be sleeping while the agent runs.
+	// Always collect on first call (lastDiskUsageUpdate is zero) or if caching is disabled.
+	cacheExtraFs := a.diskUsageCacheDuration > 0 &&
+		!a.lastDiskUsageUpdate.IsZero() &&
+		time.Since(a.lastDiskUsageUpdate) < a.diskUsageCacheDuration
+
+	// disk usage
+	for _, stats := range a.fsStats {
+		// Skip non-root filesystems if caching is active
+		if cacheExtraFs && !stats.Root {
+			continue
+		}
+		if d, err := disk.Usage(stats.Mountpoint); err == nil {
+			stats.DiskTotal = bytesToGigabytes(d.Total)
+			stats.DiskUsed = bytesToGigabytes(d.Used)
+			if stats.Root {
+				systemStats.DiskTotal = bytesToGigabytes(d.Total)
+				systemStats.DiskUsed = bytesToGigabytes(d.Used)
+				systemStats.DiskPct = twoDecimals(d.UsedPercent)
+			}
+		} else {
+			// reset stats if error (likely unmounted)
+			slog.Error("Error getting disk stats", "name", stats.Mountpoint, "err", err)
+			stats.DiskTotal = 0
+			stats.DiskUsed = 0
+			stats.TotalRead = 0
+			stats.TotalWrite = 0
+		}
+	}
+
+	// Update the last disk usage update time when we've collected extra filesystems
+	if !cacheExtraFs {
+		a.lastDiskUsageUpdate = time.Now()
+	}
+}
+
+// Updates disk I/O statistics for all monitored filesystems
+func (a *Agent) updateDiskIo(cacheTimeMs uint16, systemStats *system.Stats) {
+	// disk i/o (cache-aware per interval)
+	if ioCounters, err := disk.IOCounters(a.fsNames...); err == nil {
+		// Ensure map for this interval exists
+		if _, ok := a.diskPrev[cacheTimeMs]; !ok {
+			a.diskPrev[cacheTimeMs] = make(map[string]prevDisk)
+		}
+		now := time.Now()
+		for name, d := range ioCounters {
+			stats := a.fsStats[d.Name]
+			if stats == nil {
+				// skip devices not tracked
+				continue
+			}
+
+			// Previous snapshot for this interval and device
+			prev, hasPrev := a.diskPrev[cacheTimeMs][name]
+			if !hasPrev {
+				// Seed from agent-level fsStats if present, else seed from current
+				prev = prevDisk{readBytes: stats.TotalRead, writeBytes: stats.TotalWrite, at: stats.Time}
+				if prev.at.IsZero() {
+					prev = prevDisk{readBytes: d.ReadBytes, writeBytes: d.WriteBytes, at: now}
+				}
+			}
+
+			msElapsed := uint64(now.Sub(prev.at).Milliseconds())
+			if msElapsed < 100 {
+				// Avoid division by zero or clock issues; update snapshot and continue
+				a.diskPrev[cacheTimeMs][name] = prevDisk{readBytes: d.ReadBytes, writeBytes: d.WriteBytes, at: now}
+				continue
+			}
+
+			diskIORead := (d.ReadBytes - prev.readBytes) * 1000 / msElapsed
+			diskIOWrite := (d.WriteBytes - prev.writeBytes) * 1000 / msElapsed
+			readMbPerSecond := bytesToMegabytes(float64(diskIORead))
+			writeMbPerSecond := bytesToMegabytes(float64(diskIOWrite))
+
+			// validate values
+			if readMbPerSecond > 50_000 || writeMbPerSecond > 50_000 {
+				slog.Warn("Invalid disk I/O. Resetting.", "name", d.Name, "read", readMbPerSecond, "write", writeMbPerSecond)
+				// Reset interval snapshot and seed from current
+				a.diskPrev[cacheTimeMs][name] = prevDisk{readBytes: d.ReadBytes, writeBytes: d.WriteBytes, at: now}
+				// also refresh agent baseline to avoid future negatives
+				a.initializeDiskIoStats(ioCounters)
+				continue
+			}
+
+			// Update per-interval snapshot
+			a.diskPrev[cacheTimeMs][name] = prevDisk{readBytes: d.ReadBytes, writeBytes: d.WriteBytes, at: now}
+
+			// Update global fsStats baseline for cross-interval correctness
+			stats.Time = now
+			stats.TotalRead = d.ReadBytes
+			stats.TotalWrite = d.WriteBytes
+			stats.DiskReadPs = readMbPerSecond
+			stats.DiskWritePs = writeMbPerSecond
+			stats.DiskReadBytes = diskIORead
+			stats.DiskWriteBytes = diskIOWrite
+
+			if stats.Root {
+				systemStats.DiskReadPs = stats.DiskReadPs
+				systemStats.DiskWritePs = stats.DiskWritePs
+				systemStats.DiskIO[0] = diskIORead
+				systemStats.DiskIO[1] = diskIOWrite
+			}
+		}
+	}
+}
+
+// getRootMountPoint returns the appropriate root mount point for the system
+// For immutable systems like Fedora Silverblue, it returns /sysroot instead of /
+func (a *Agent) getRootMountPoint() string {
+	// 1. Check if /etc/os-release contains indicators of an immutable system
+	if osReleaseContent, err := os.ReadFile("/etc/os-release"); err == nil {
+		content := string(osReleaseContent)
+		if strings.Contains(content, "fedora") && strings.Contains(content, "silverblue") ||
+			strings.Contains(content, "coreos") ||
+			strings.Contains(content, "flatcar") ||
+			strings.Contains(content, "rhel-atomic") ||
+			strings.Contains(content, "centos-atomic") {
+			// Verify that /sysroot exists before returning it
+			if _, err := os.Stat("/sysroot"); err == nil {
+				return "/sysroot"
+			}
+		}
+	}
+
+	// 2. Check if /run/ostree is present (ostree-based systems like Silverblue)
+	if _, err := os.Stat("/run/ostree"); err == nil {
+		// Verify that /sysroot exists before returning it
+		if _, err := os.Stat("/sysroot"); err == nil {
+			return "/sysroot"
+		}
+	}
+
+	return "/"
+}

agent/disk_test.go

@@ -0,0 +1,319 @@
+//go:build testing
+// +build testing
+
+package agent
+
+import (
+	"os"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/henrygd/beszel/internal/entities/system"
+	"github.com/shirou/gopsutil/v4/disk"
+	"github.com/stretchr/testify/assert"
+)
+
+func TestParseFilesystemEntry(t *testing.T) {
+	tests := []struct {
+		name         string
+		input        string
+		expectedFs   string
+		expectedName string
+	}{
+		{
+			name:         "simple device name",
+			input:        "sda1",
+			expectedFs:   "sda1",
+			expectedName: "",
+		},
+		{
+			name:         "device with custom name",
+			input:        "sda1__my-storage",
+			expectedFs:   "sda1",
+			expectedName: "my-storage",
+		},
+		{
+			name:         "full device path with custom name",
+			input:        "/dev/sdb1__backup-drive",
+			expectedFs:   "/dev/sdb1",
+			expectedName: "backup-drive",
+		},
+		{
+			name:         "NVMe device with custom name",
+			input:        "nvme0n1p2__fast-ssd",
+			expectedFs:   "nvme0n1p2",
+			expectedName: "fast-ssd",
+		},
+		{
+			name:         "whitespace trimmed",
+			input:        "  sda2__trimmed-name  ",
+			expectedFs:   "sda2",
+			expectedName: "trimmed-name",
+		},
+		{
+			name:         "empty custom name",
+			input:        "sda3__",
+			expectedFs:   "sda3",
+			expectedName: "",
+		},
+		{
+			name:         "empty device name",
+			input:        "__just-custom",
+			expectedFs:   "",
+			expectedName: "just-custom",
+		},
+		{
+			name:         "multiple underscores in custom name",
+			input:        "sda1__my_custom_drive",
+			expectedFs:   "sda1",
+			expectedName: "my_custom_drive",
+		},
+		{
+			name:         "custom name with spaces",
+			input:        "sda1__My Storage Drive",
+			expectedFs:   "sda1",
+			expectedName: "My Storage Drive",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			fsEntry := strings.TrimSpace(tt.input)
+			var fs, customName string
+			if parts := strings.SplitN(fsEntry, "__", 2); len(parts) == 2 {
+				fs = strings.TrimSpace(parts[0])
+				customName = strings.TrimSpace(parts[1])
+			} else {
+				fs = fsEntry
+			}
+
+			assert.Equal(t, tt.expectedFs, fs)
+			assert.Equal(t, tt.expectedName, customName)
+		})
+	}
+}
+
+func TestInitializeDiskInfoWithCustomNames(t *testing.T) {
+	// Set up environment variables
+	oldEnv := os.Getenv("EXTRA_FILESYSTEMS")
+	defer func() {
+		if oldEnv != "" {
+			os.Setenv("EXTRA_FILESYSTEMS", oldEnv)
+		} else {
+			os.Unsetenv("EXTRA_FILESYSTEMS")
+		}
+	}()
+
+	// Test with custom names
+	os.Setenv("EXTRA_FILESYSTEMS", "sda1__my-storage,/dev/sdb1__backup-drive,nvme0n1p2")
+
+	// Mock disk partitions (we'll just test the parsing logic)
+	// Since the actual disk operations are system-dependent, we'll focus on the parsing
+	testCases := []struct {
+		envValue      string
+		expectedFs    []string
+		expectedNames map[string]string
+	}{
+		{
+			envValue:   "sda1__my-storage,sdb1__backup-drive",
+			expectedFs: []string{"sda1", "sdb1"},
+			expectedNames: map[string]string{
+				"sda1": "my-storage",
+				"sdb1": "backup-drive",
+			},
+		},
+		{
+			envValue:   "sda1,nvme0n1p2__fast-ssd",
+			expectedFs: []string{"sda1", "nvme0n1p2"},
+			expectedNames: map[string]string{
+				"nvme0n1p2": "fast-ssd",
+			},
+		},
+	}
+
+	for _, tc := range testCases {
+		t.Run("env_"+tc.envValue, func(t *testing.T) {
+			os.Setenv("EXTRA_FILESYSTEMS", tc.envValue)
+
+			// Create mock partitions that would match our test cases
+			partitions := []disk.PartitionStat{}
+			for _, fs := range tc.expectedFs {
+				if strings.HasPrefix(fs, "/dev/") {
+					partitions = append(partitions, disk.PartitionStat{
+						Device:     fs,
+						Mountpoint: fs,
+					})
+				} else {
+					partitions = append(partitions, disk.PartitionStat{
+						Device:     "/dev/" + fs,
+						Mountpoint: "/" + fs,
+					})
+				}
+			}
+
+			// Test the parsing logic by calling the relevant part
+			// We'll create a simplified version to test just the parsing
+			extraFilesystems := tc.envValue
+			for _, fsEntry := range strings.Split(extraFilesystems, ",") {
+				// Parse the entry
+				fsEntry = strings.TrimSpace(fsEntry)
+				var fs, customName string
+				if parts := strings.SplitN(fsEntry, "__", 2); len(parts) == 2 {
+					fs = strings.TrimSpace(parts[0])
+					customName = strings.TrimSpace(parts[1])
+				} else {
+					fs = fsEntry
+				}
+
+				// Verify the device is in our expected list
+				assert.Contains(t, tc.expectedFs, fs, "parsed device should be in expected list")
+
+				// Check if custom name should exist
+				if expectedName, exists := tc.expectedNames[fs]; exists {
+					assert.Equal(t, expectedName, customName, "custom name should match expected")
+				} else {
+					assert.Empty(t, customName, "custom name should be empty when not expected")
+				}
+			}
+		})
+	}
+}
+
+func TestFsStatsWithCustomNames(t *testing.T) {
+	// Test that FsStats properly stores custom names
+	fsStats := &system.FsStats{
+		Mountpoint: "/mnt/storage",
+		Name:       "my-custom-storage",
+		DiskTotal:  100.0,
+		DiskUsed:   50.0,
+	}
+
+	assert.Equal(t, "my-custom-storage", fsStats.Name)
+	assert.Equal(t, "/mnt/storage", fsStats.Mountpoint)
+	assert.Equal(t, 100.0, fsStats.DiskTotal)
+	assert.Equal(t, 50.0, fsStats.DiskUsed)
+}
+
+func TestExtraFsKeyGeneration(t *testing.T) {
+	// Test the logic for generating ExtraFs keys with custom names
+	testCases := []struct {
+		name        string
+		deviceName  string
+		customName  string
+		expectedKey string
+	}{
+		{
+			name:        "with custom name",
+			deviceName:  "sda1",
+			customName:  "my-storage",
+			expectedKey: "my-storage",
+		},
+		{
+			name:        "without custom name",
+			deviceName:  "sda1",
+			customName:  "",
+			expectedKey: "sda1",
+		},
+		{
+			name:        "empty custom name falls back to device",
+			deviceName:  "nvme0n1p2",
+			customName:  "",
+			expectedKey: "nvme0n1p2",
+		},
+	}
+
+	for _, tc := range testCases {
+		t.Run(tc.name, func(t *testing.T) {
+			// Simulate the key generation logic from agent.go
+			key := tc.deviceName
+			if tc.customName != "" {
+				key = tc.customName
+			}
+			assert.Equal(t, tc.expectedKey, key)
+		})
+	}
+}
+
+func TestDiskUsageCaching(t *testing.T) {
+	t.Run("caching disabled updates all filesystems", func(t *testing.T) {
+		agent := &Agent{
+			fsStats: map[string]*system.FsStats{
+				"sda": {Root: true, Mountpoint: "/"},
+				"sdb": {Root: false, Mountpoint: "/mnt/storage"},
+			},
+			diskUsageCacheDuration: 0, // caching disabled
+		}
+
+		var stats system.Stats
+		agent.updateDiskUsage(&stats)
+
+		// Both should be updated (non-zero values from disk.Usage)
+		// Root stats should be populated in systemStats
+		assert.True(t, agent.lastDiskUsageUpdate.IsZero() || !agent.lastDiskUsageUpdate.IsZero(),
+			"lastDiskUsageUpdate should be set when caching is disabled")
+	})
+
+	t.Run("caching enabled always updates root filesystem", func(t *testing.T) {
+		agent := &Agent{
+			fsStats: map[string]*system.FsStats{
+				"sda": {Root: true, Mountpoint: "/", DiskTotal: 100, DiskUsed: 50},
+				"sdb": {Root: false, Mountpoint: "/mnt/storage", DiskTotal: 200, DiskUsed: 100},
+			},
+			diskUsageCacheDuration: 1 * time.Hour,
+			lastDiskUsageUpdate:    time.Now(), // cache is fresh
+		}
+
+		// Store original extra fs values
+		originalExtraTotal := agent.fsStats["sdb"].DiskTotal
+		originalExtraUsed := agent.fsStats["sdb"].DiskUsed
+
+		var stats system.Stats
+		agent.updateDiskUsage(&stats)
+
+		// Root should be updated (systemStats populated from disk.Usage call)
+		// We can't easily check if disk.Usage was called, but we verify the flow works
+
+		// Extra filesystem should retain cached values (not reset)
+		assert.Equal(t, originalExtraTotal, agent.fsStats["sdb"].DiskTotal,
+			"extra filesystem DiskTotal should be unchanged when cached")
+		assert.Equal(t, originalExtraUsed, agent.fsStats["sdb"].DiskUsed,
+			"extra filesystem DiskUsed should be unchanged when cached")
+	})
+
+	t.Run("first call always updates all filesystems", func(t *testing.T) {
+		agent := &Agent{
+			fsStats: map[string]*system.FsStats{
+				"sda": {Root: true, Mountpoint: "/"},
+				"sdb": {Root: false, Mountpoint: "/mnt/storage"},
+			},
+			diskUsageCacheDuration: 1 * time.Hour,
+			// lastDiskUsageUpdate is zero (first call)
+		}
+
+		var stats system.Stats
+		agent.updateDiskUsage(&stats)
+
+		// After first call, lastDiskUsageUpdate should be set
+		assert.False(t, agent.lastDiskUsageUpdate.IsZero(),
+			"lastDiskUsageUpdate should be set after first call")
+	})
+
+	t.Run("expired cache updates extra filesystems", func(t *testing.T) {
+		agent := &Agent{
+			fsStats: map[string]*system.FsStats{
+				"sda": {Root: true, Mountpoint: "/"},
+				"sdb": {Root: false, Mountpoint: "/mnt/storage"},
+			},
+			diskUsageCacheDuration: 1 * time.Millisecond,
+			lastDiskUsageUpdate:    time.Now().Add(-1 * time.Second), // cache expired
+		}
+
+		var stats system.Stats
+		agent.updateDiskUsage(&stats)
+
+		// lastDiskUsageUpdate should be refreshed since cache expired
+		assert.True(t, time.Since(agent.lastDiskUsageUpdate) < time.Second,
+			"lastDiskUsageUpdate should be refreshed when cache expires")
+	})
+}

agent/docker.go

@@ -0,0 +1,773 @@
+package agent
+
+import (
+	"bytes"
+	"context"
+	"encoding/binary"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"io"
+	"log/slog"
+	"net"
+	"net/http"
+	"net/url"
+	"os"
+	"path"
+	"regexp"
+	"strings"
+	"sync"
+	"time"
+
+	"github.com/henrygd/beszel/agent/deltatracker"
+	"github.com/henrygd/beszel/internal/entities/container"
+
+	"github.com/blang/semver"
+)
+
+// ansiEscapePattern matches ANSI escape sequences (colors, cursor movement, etc.)
+// This includes CSI sequences like \x1b[...m and simple escapes like \x1b[K
+var ansiEscapePattern = regexp.MustCompile(`\x1b\[[0-9;]*[a-zA-Z]|\x1b\][^\x07]*\x07|\x1b[@-Z\\-_]`)
+
+const (
+	// Docker API timeout in milliseconds
+	dockerTimeoutMs = 2100
+	// Maximum realistic network speed (5 GB/s) to detect bad deltas
+	maxNetworkSpeedBps uint64 = 5e9
+	// Maximum conceivable memory usage of a container (100TB) to detect bad memory stats
+	maxMemoryUsage uint64 = 100 * 1024 * 1024 * 1024 * 1024
+	// Number of log lines to request when fetching container logs
+	dockerLogsTail = 200
+	// Maximum size of a single log frame (1MB) to prevent memory exhaustion
+	// A single log line larger than 1MB is likely an error or misconfiguration
+	maxLogFrameSize = 1024 * 1024
+	// Maximum total log content size (5MB) to prevent memory exhaustion
+	// This provides a reasonable limit for network transfer and browser rendering
+	maxTotalLogSize = 5 * 1024 * 1024
+)
+
+type dockerManager struct {
+	client              *http.Client                // Client to query Docker API
+	wg                  sync.WaitGroup              // WaitGroup to wait for all goroutines to finish
+	sem                 chan struct{}               // Semaphore to limit concurrent container requests
+	containerStatsMutex sync.RWMutex                // Mutex to prevent concurrent access to containerStatsMap
+	apiContainerList    []*container.ApiInfo        // List of containers from Docker API
+	containerStatsMap   map[string]*container.Stats // Keeps track of container stats
+	validIds            map[string]struct{}         // Map of valid container ids, used to prune invalid containers from containerStatsMap
+	goodDockerVersion   bool                        // Whether docker version is at least 25.0.0 (one-shot works correctly)
+	isWindows           bool                        // Whether the Docker Engine API is running on Windows
+	buf                 *bytes.Buffer               // Buffer to store and read response bodies
+	decoder             *json.Decoder               // Reusable JSON decoder that reads from buf
+	apiStats            *container.ApiStats         // Reusable API stats object
+	excludeContainers   []string                    // Patterns to exclude containers by name
+	usingPodman         bool                        // Whether the Docker Engine API is running on Podman
+
+	// Cache-time-aware tracking for CPU stats (similar to cpu.go)
+	// Maps cache time intervals to container-specific CPU usage tracking
+	lastCpuContainer map[uint16]map[string]uint64    // cacheTimeMs -> containerId -> last cpu container usage
+	lastCpuSystem    map[uint16]map[string]uint64    // cacheTimeMs -> containerId -> last cpu system usage
+	lastCpuReadTime  map[uint16]map[string]time.Time // cacheTimeMs -> containerId -> last read time (Windows)
+
+	// Network delta trackers - one per cache time to avoid interference
+	// cacheTimeMs -> DeltaTracker for network bytes sent/received
+	networkSentTrackers map[uint16]*deltatracker.DeltaTracker[string, uint64]
+	networkRecvTrackers map[uint16]*deltatracker.DeltaTracker[string, uint64]
+}
+
+// userAgentRoundTripper is a custom http.RoundTripper that adds a User-Agent header to all requests
+type userAgentRoundTripper struct {
+	rt        http.RoundTripper
+	userAgent string
+}
+
+// RoundTrip implements the http.RoundTripper interface
+func (u *userAgentRoundTripper) RoundTrip(req *http.Request) (*http.Response, error) {
+	req.Header.Set("User-Agent", u.userAgent)
+	return u.rt.RoundTrip(req)
+}
+
+// Add goroutine to the queue
+func (d *dockerManager) queue() {
+	d.wg.Add(1)
+	if d.goodDockerVersion {
+		d.sem <- struct{}{}
+	}
+}
+
+// Remove goroutine from the queue
+func (d *dockerManager) dequeue() {
+	d.wg.Done()
+	if d.goodDockerVersion {
+		<-d.sem
+	}
+}
+
+// shouldExcludeContainer checks if a container name matches any exclusion pattern
+func (dm *dockerManager) shouldExcludeContainer(name string) bool {
+	if len(dm.excludeContainers) == 0 {
+		return false
+	}
+	for _, pattern := range dm.excludeContainers {
+		if match, _ := path.Match(pattern, name); match {
+			return true
+		}
+	}
+	return false
+}
+
+// Returns stats for all running containers with cache-time-aware delta tracking
+func (dm *dockerManager) getDockerStats(cacheTimeMs uint16) ([]*container.Stats, error) {
+	resp, err := dm.client.Get("http://localhost/containers/json")
+	if err != nil {
+		return nil, err
+	}
+
+	dm.apiContainerList = dm.apiContainerList[:0]
+	if err := dm.decode(resp, &dm.apiContainerList); err != nil {
+		return nil, err
+	}
+
+	dm.isWindows = strings.Contains(resp.Header.Get("Server"), "windows")
+
+	containersLength := len(dm.apiContainerList)
+
+	// store valid ids to clean up old container ids from map
+	if dm.validIds == nil {
+		dm.validIds = make(map[string]struct{}, containersLength)
+	} else {
+		clear(dm.validIds)
+	}
+
+	var failedContainers []*container.ApiInfo
+
+	for _, ctr := range dm.apiContainerList {
+		ctr.IdShort = ctr.Id[:12]
+
+		// Skip this container if it matches the exclusion pattern
+		if dm.shouldExcludeContainer(ctr.Names[0][1:]) {
+			slog.Debug("Excluding container", "name", ctr.Names[0][1:])
+			continue
+		}
+
+		dm.validIds[ctr.IdShort] = struct{}{}
+		// check if container is less than 1 minute old (possible restart)
+		// note: can't use Created field because it's not updated on restart
+		if strings.Contains(ctr.Status, "second") {
+			// if so, remove old container data
+			dm.deleteContainerStatsSync(ctr.IdShort)
+		}
+		dm.queue()
+		go func(ctr *container.ApiInfo) {
+			defer dm.dequeue()
+			err := dm.updateContainerStats(ctr, cacheTimeMs)
+			// if error, delete from map and add to failed list to retry
+			if err != nil {
+				dm.containerStatsMutex.Lock()
+				delete(dm.containerStatsMap, ctr.IdShort)
+				failedContainers = append(failedContainers, ctr)
+				dm.containerStatsMutex.Unlock()
+			}
+		}(ctr)
+	}
+
+	dm.wg.Wait()
+
+	// retry failed containers separately so we can run them in parallel (docker 24 bug)
+	if len(failedContainers) > 0 {
+		slog.Debug("Retrying failed containers", "count", len(failedContainers))
+		for i := range failedContainers {
+			ctr := failedContainers[i]
+			dm.queue()
+			go func(ctr *container.ApiInfo) {
+				defer dm.dequeue()
+				if err2 := dm.updateContainerStats(ctr, cacheTimeMs); err2 != nil {
+					slog.Error("Error getting container stats", "err", err2)
+				}
+			}(ctr)
+		}
+		dm.wg.Wait()
+	}
+
+	// populate final stats and remove old / invalid container stats
+	stats := make([]*container.Stats, 0, containersLength)
+	for id, v := range dm.containerStatsMap {
+		if _, exists := dm.validIds[id]; !exists {
+			delete(dm.containerStatsMap, id)
+		} else {
+			stats = append(stats, v)
+		}
+	}
+
+	// prepare network trackers for next interval for this cache time
+	dm.cycleNetworkDeltasForCacheTime(cacheTimeMs)
+
+	return stats, nil
+}
+
+// initializeCpuTracking initializes CPU tracking maps for a specific cache time interval
+func (dm *dockerManager) initializeCpuTracking(cacheTimeMs uint16) {
+	// Initialize cache time maps if they don't exist
+	if dm.lastCpuContainer[cacheTimeMs] == nil {
+		dm.lastCpuContainer[cacheTimeMs] = make(map[string]uint64)
+	}
+	if dm.lastCpuSystem[cacheTimeMs] == nil {
+		dm.lastCpuSystem[cacheTimeMs] = make(map[string]uint64)
+	}
+	// Ensure the outer map exists before indexing
+	if dm.lastCpuReadTime == nil {
+		dm.lastCpuReadTime = make(map[uint16]map[string]time.Time)
+	}
+	if dm.lastCpuReadTime[cacheTimeMs] == nil {
+		dm.lastCpuReadTime[cacheTimeMs] = make(map[string]time.Time)
+	}
+}
+
+// getCpuPreviousValues returns previous CPU values for a container and cache time interval
+func (dm *dockerManager) getCpuPreviousValues(cacheTimeMs uint16, containerId string) (uint64, uint64) {
+	return dm.lastCpuContainer[cacheTimeMs][containerId], dm.lastCpuSystem[cacheTimeMs][containerId]
+}
+
+// setCpuCurrentValues stores current CPU values for a container and cache time interval
+func (dm *dockerManager) setCpuCurrentValues(cacheTimeMs uint16, containerId string, cpuContainer, cpuSystem uint64) {
+	dm.lastCpuContainer[cacheTimeMs][containerId] = cpuContainer
+	dm.lastCpuSystem[cacheTimeMs][containerId] = cpuSystem
+}
+
+// calculateMemoryUsage calculates memory usage from Docker API stats
+func calculateMemoryUsage(apiStats *container.ApiStats, isWindows bool) (uint64, error) {
+	if isWindows {
+		return apiStats.MemoryStats.PrivateWorkingSet, nil
+	}
+
+	memCache := apiStats.MemoryStats.Stats.InactiveFile
+	if memCache == 0 {
+		memCache = apiStats.MemoryStats.Stats.Cache
+	}
+
+	usedDelta := apiStats.MemoryStats.Usage - memCache
+	if usedDelta <= 0 || usedDelta > maxMemoryUsage {
+		return 0, fmt.Errorf("bad memory stats")
+	}
+
+	return usedDelta, nil
+}
+
+// getNetworkTracker returns the DeltaTracker for a specific cache time, creating it if needed
+func (dm *dockerManager) getNetworkTracker(cacheTimeMs uint16, isSent bool) *deltatracker.DeltaTracker[string, uint64] {
+	var trackers map[uint16]*deltatracker.DeltaTracker[string, uint64]
+	if isSent {
+		trackers = dm.networkSentTrackers
+	} else {
+		trackers = dm.networkRecvTrackers
+	}
+
+	if trackers[cacheTimeMs] == nil {
+		trackers[cacheTimeMs] = deltatracker.NewDeltaTracker[string, uint64]()
+	}
+
+	return trackers[cacheTimeMs]
+}
+
+// cycleNetworkDeltasForCacheTime cycles the network delta trackers for a specific cache time
+func (dm *dockerManager) cycleNetworkDeltasForCacheTime(cacheTimeMs uint16) {
+	if dm.networkSentTrackers[cacheTimeMs] != nil {
+		dm.networkSentTrackers[cacheTimeMs].Cycle()
+	}
+	if dm.networkRecvTrackers[cacheTimeMs] != nil {
+		dm.networkRecvTrackers[cacheTimeMs].Cycle()
+	}
+}
+
+// calculateNetworkStats calculates network sent/receive deltas using DeltaTracker
+func (dm *dockerManager) calculateNetworkStats(ctr *container.ApiInfo, apiStats *container.ApiStats, stats *container.Stats, initialized bool, name string, cacheTimeMs uint16) (uint64, uint64) {
+	var total_sent, total_recv uint64
+	for _, v := range apiStats.Networks {
+		total_sent += v.TxBytes
+		total_recv += v.RxBytes
+	}
+
+	// Get the DeltaTracker for this specific cache time
+	sentTracker := dm.getNetworkTracker(cacheTimeMs, true)
+	recvTracker := dm.getNetworkTracker(cacheTimeMs, false)
+
+	// Set current values in the cache-time-specific DeltaTracker
+	sentTracker.Set(ctr.IdShort, total_sent)
+	recvTracker.Set(ctr.IdShort, total_recv)
+
+	// Get deltas (bytes since last measurement)
+	sent_delta_raw := sentTracker.Delta(ctr.IdShort)
+	recv_delta_raw := recvTracker.Delta(ctr.IdShort)
+
+	// Calculate bytes per second independently for Tx and Rx if we have previous data
+	var sent_delta, recv_delta uint64
+	if initialized {
+		millisecondsElapsed := uint64(time.Since(stats.PrevReadTime).Milliseconds())
+		if millisecondsElapsed > 0 {
+			if sent_delta_raw > 0 {
+				sent_delta = sent_delta_raw * 1000 / millisecondsElapsed
+				if sent_delta > maxNetworkSpeedBps {
+					slog.Warn("Bad network delta", "container", name)
+					sent_delta = 0
+				}
+			}
+			if recv_delta_raw > 0 {
+				recv_delta = recv_delta_raw * 1000 / millisecondsElapsed
+				if recv_delta > maxNetworkSpeedBps {
+					slog.Warn("Bad network delta", "container", name)
+					recv_delta = 0
+				}
+			}
+		}
+	}
+
+	return sent_delta, recv_delta
+}
+
+// validateCpuPercentage checks if CPU percentage is within valid range
+func validateCpuPercentage(cpuPct float64, containerName string) error {
+	if cpuPct > 100 {
+		return fmt.Errorf("%s cpu pct greater than 100: %+v", containerName, cpuPct)
+	}
+	return nil
+}
+
+// updateContainerStatsValues updates the final stats values
+func updateContainerStatsValues(stats *container.Stats, cpuPct float64, usedMemory uint64, sent_delta, recv_delta uint64, readTime time.Time) {
+	stats.Cpu = twoDecimals(cpuPct)
+	stats.Mem = bytesToMegabytes(float64(usedMemory))
+	stats.NetworkSent = bytesToMegabytes(float64(sent_delta))
+	stats.NetworkRecv = bytesToMegabytes(float64(recv_delta))
+	stats.PrevReadTime = readTime
+}
+
+func parseDockerStatus(status string) (string, container.DockerHealth) {
+	trimmed := strings.TrimSpace(status)
+	if trimmed == "" {
+		return "", container.DockerHealthNone
+	}
+
+	// Remove "About " from status
+	trimmed = strings.Replace(trimmed, "About ", "", 1)
+
+	openIdx := strings.LastIndex(trimmed, "(")
+	if openIdx == -1 || !strings.HasSuffix(trimmed, ")") {
+		return trimmed, container.DockerHealthNone
+	}
+
+	statusText := strings.TrimSpace(trimmed[:openIdx])
+	if statusText == "" {
+		statusText = trimmed
+	}
+
+	healthText := strings.ToLower(strings.TrimSpace(strings.TrimSuffix(trimmed[openIdx+1:], ")")))
+	// Some Docker statuses include a "health:" prefix inside the parentheses.
+	// Strip it so it maps correctly to the known health states.
+	if colonIdx := strings.IndexRune(healthText, ':'); colonIdx != -1 {
+		prefix := strings.TrimSpace(healthText[:colonIdx])
+		if prefix == "health" || prefix == "health status" {
+			healthText = strings.TrimSpace(healthText[colonIdx+1:])
+		}
+	}
+	if health, ok := container.DockerHealthStrings[healthText]; ok {
+		return statusText, health
+	}
+
+	return trimmed, container.DockerHealthNone
+}
+
+// Updates stats for individual container with cache-time-aware delta tracking
+func (dm *dockerManager) updateContainerStats(ctr *container.ApiInfo, cacheTimeMs uint16) error {
+	name := ctr.Names[0][1:]
+
+	resp, err := dm.client.Get(fmt.Sprintf("http://localhost/containers/%s/stats?stream=0&one-shot=1", ctr.IdShort))
+	if err != nil {
+		return err
+	}
+
+	dm.containerStatsMutex.Lock()
+	defer dm.containerStatsMutex.Unlock()
+
+	// add empty values if they doesn't exist in map
+	stats, initialized := dm.containerStatsMap[ctr.IdShort]
+	if !initialized {
+		stats = &container.Stats{Name: name, Id: ctr.IdShort, Image: ctr.Image}
+		dm.containerStatsMap[ctr.IdShort] = stats
+	}
+
+	stats.Id = ctr.IdShort
+
+	statusText, health := parseDockerStatus(ctr.Status)
+	stats.Status = statusText
+	stats.Health = health
+
+	// reset current stats
+	stats.Cpu = 0
+	stats.Mem = 0
+	stats.NetworkSent = 0
+	stats.NetworkRecv = 0
+
+	res := dm.apiStats
+	res.Networks = nil
+	if err := dm.decode(resp, res); err != nil {
+		return err
+	}
+
+	// Initialize CPU tracking for this cache time interval
+	dm.initializeCpuTracking(cacheTimeMs)
+
+	// Get previous CPU values
+	prevCpuContainer, prevCpuSystem := dm.getCpuPreviousValues(cacheTimeMs, ctr.IdShort)
+
+	// Calculate CPU percentage based on platform
+	var cpuPct float64
+	if dm.isWindows {
+		prevRead := dm.lastCpuReadTime[cacheTimeMs][ctr.IdShort]
+		cpuPct = res.CalculateCpuPercentWindows(prevCpuContainer, prevRead)
+	} else {
+		cpuPct = res.CalculateCpuPercentLinux(prevCpuContainer, prevCpuSystem)
+	}
+
+	// Calculate memory usage
+	usedMemory, err := calculateMemoryUsage(res, dm.isWindows)
+	if err != nil {
+		return fmt.Errorf("%s - %w - see https://github.com/henrygd/beszel/issues/144", name, err)
+	}
+
+	// Store current CPU stats for next calculation
+	currentCpuContainer := res.CPUStats.CPUUsage.TotalUsage
+	currentCpuSystem := res.CPUStats.SystemUsage
+	dm.setCpuCurrentValues(cacheTimeMs, ctr.IdShort, currentCpuContainer, currentCpuSystem)
+
+	// Validate CPU percentage
+	if err := validateCpuPercentage(cpuPct, name); err != nil {
+		return err
+	}
+
+	// Calculate network stats using DeltaTracker
+	sent_delta, recv_delta := dm.calculateNetworkStats(ctr, res, stats, initialized, name, cacheTimeMs)
+
+	// Store current network values for legacy compatibility
+	var total_sent, total_recv uint64
+	for _, v := range res.Networks {
+		total_sent += v.TxBytes
+		total_recv += v.RxBytes
+	}
+	stats.PrevNet.Sent, stats.PrevNet.Recv = total_sent, total_recv
+
+	// Update final stats values
+	updateContainerStatsValues(stats, cpuPct, usedMemory, sent_delta, recv_delta, res.Read)
+	// store per-cache-time read time for Windows CPU percent calc
+	dm.lastCpuReadTime[cacheTimeMs][ctr.IdShort] = res.Read
+
+	return nil
+}
+
+// Delete container stats from map using mutex
+func (dm *dockerManager) deleteContainerStatsSync(id string) {
+	dm.containerStatsMutex.Lock()
+	defer dm.containerStatsMutex.Unlock()
+	delete(dm.containerStatsMap, id)
+	for ct := range dm.lastCpuContainer {
+		delete(dm.lastCpuContainer[ct], id)
+	}
+	for ct := range dm.lastCpuSystem {
+		delete(dm.lastCpuSystem[ct], id)
+	}
+	for ct := range dm.lastCpuReadTime {
+		delete(dm.lastCpuReadTime[ct], id)
+	}
+}
+
+// Creates a new http client for Docker or Podman API
+func newDockerManager() *dockerManager {
+	dockerHost, exists := GetEnv("DOCKER_HOST")
+	if exists {
+		// return nil if set to empty string
+		if dockerHost == "" {
+			return nil
+		}
+	} else {
+		dockerHost = getDockerHost()
+	}
+
+	parsedURL, err := url.Parse(dockerHost)
+	if err != nil {
+		os.Exit(1)
+	}
+
+	transport := &http.Transport{
+		DisableCompression: true,
+		MaxConnsPerHost:    0,
+	}
+
+	switch parsedURL.Scheme {
+	case "unix":
+		transport.DialContext = func(ctx context.Context, proto, addr string) (net.Conn, error) {
+			return (&net.Dialer{}).DialContext(ctx, "unix", parsedURL.Path)
+		}
+	case "tcp", "http", "https":
+		transport.DialContext = func(ctx context.Context, proto, addr string) (net.Conn, error) {
+			return (&net.Dialer{}).DialContext(ctx, "tcp", parsedURL.Host)
+		}
+	default:
+		slog.Error("Invalid DOCKER_HOST", "scheme", parsedURL.Scheme)
+		os.Exit(1)
+	}
+
+	// configurable timeout
+	timeout := time.Millisecond * time.Duration(dockerTimeoutMs)
+	if t, set := GetEnv("DOCKER_TIMEOUT"); set {
+		timeout, err = time.ParseDuration(t)
+		if err != nil {
+			slog.Error(err.Error())
+			os.Exit(1)
+		}
+		slog.Info("DOCKER_TIMEOUT", "timeout", timeout)
+	}
+
+	// Custom user-agent to avoid docker bug: https://github.com/docker/for-mac/issues/7575
+	userAgentTransport := &userAgentRoundTripper{
+		rt:        transport,
+		userAgent: "Docker-Client/",
+	}
+
+	// Read container exclusion patterns from environment variable
+	var excludeContainers []string
+	if excludeStr, set := GetEnv("EXCLUDE_CONTAINERS"); set && excludeStr != "" {
+		parts := strings.SplitSeq(excludeStr, ",")
+		for part := range parts {
+			trimmed := strings.TrimSpace(part)
+			if trimmed != "" {
+				excludeContainers = append(excludeContainers, trimmed)
+			}
+		}
+		slog.Info("EXCLUDE_CONTAINERS", "patterns", excludeContainers)
+	}
+
+	manager := &dockerManager{
+		client: &http.Client{
+			Timeout:   timeout,
+			Transport: userAgentTransport,
+		},
+		containerStatsMap: make(map[string]*container.Stats),
+		sem:               make(chan struct{}, 5),
+		apiContainerList:  []*container.ApiInfo{},
+		apiStats:          &container.ApiStats{},
+		excludeContainers: excludeContainers,
+
+		// Initialize cache-time-aware tracking structures
+		lastCpuContainer:    make(map[uint16]map[string]uint64),
+		lastCpuSystem:       make(map[uint16]map[string]uint64),
+		lastCpuReadTime:     make(map[uint16]map[string]time.Time),
+		networkSentTrackers: make(map[uint16]*deltatracker.DeltaTracker[string, uint64]),
+		networkRecvTrackers: make(map[uint16]*deltatracker.DeltaTracker[string, uint64]),
+	}
+
+	// If using podman, return client
+	if strings.Contains(dockerHost, "podman") {
+		manager.usingPodman = true
+		manager.goodDockerVersion = true
+		return manager
+	}
+
+	// this can take up to 5 seconds with retry, so run in goroutine
+	go manager.checkDockerVersion()
+
+	// give version check a chance to complete before returning
+	time.Sleep(50 * time.Millisecond)
+
+	return manager
+}
+
+// checkDockerVersion checks Docker version and sets goodDockerVersion if at least 25.0.0.
+// Versions before 25.0.0 have a bug with one-shot which requires all requests to be made in one batch.
+func (dm *dockerManager) checkDockerVersion() {
+	var err error
+	var resp *http.Response
+	var versionInfo struct {
+		Version string `json:"Version"`
+	}
+	const versionMaxTries = 2
+	for i := 1; i <= versionMaxTries; i++ {
+		resp, err = dm.client.Get("http://localhost/version")
+		if err == nil {
+			break
+		}
+		if resp != nil {
+			resp.Body.Close()
+		}
+		if i < versionMaxTries {
+			slog.Debug("Failed to get Docker version; retrying", "attempt", i, "error", err)
+			time.Sleep(5 * time.Second)
+		}
+	}
+	if err != nil {
+		return
+	}
+	if err := dm.decode(resp, &versionInfo); err != nil {
+		return
+	}
+	// if version > 24, one-shot works correctly and we can limit concurrent operations
+	if dockerVersion, err := semver.Parse(versionInfo.Version); err == nil && dockerVersion.Major > 24 {
+		dm.goodDockerVersion = true
+	} else {
+		slog.Info(fmt.Sprintf("Docker %s is outdated. Upgrade if possible. See https://github.com/henrygd/beszel/issues/58", versionInfo.Version))
+	}
+}
+
+// Decodes Docker API JSON response using a reusable buffer and decoder. Not thread safe.
+func (dm *dockerManager) decode(resp *http.Response, d any) error {
+	if dm.buf == nil {
+		// initialize buffer with 256kb starting size
+		dm.buf = bytes.NewBuffer(make([]byte, 0, 1024*256))
+		dm.decoder = json.NewDecoder(dm.buf)
+	}
+	defer resp.Body.Close()
+	defer dm.buf.Reset()
+	_, err := dm.buf.ReadFrom(resp.Body)
+	if err != nil {
+		return err
+	}
+	return dm.decoder.Decode(d)
+}
+
+// Test docker / podman sockets and return if one exists
+func getDockerHost() string {
+	scheme := "unix://"
+	socks := []string{"/var/run/docker.sock", fmt.Sprintf("/run/user/%v/podman/podman.sock", os.Getuid())}
+	for _, sock := range socks {
+		if _, err := os.Stat(sock); err == nil {
+			return scheme + sock
+		}
+	}
+	return scheme + socks[0]
+}
+
+// getContainerInfo fetches the inspection data for a container
+func (dm *dockerManager) getContainerInfo(ctx context.Context, containerID string) ([]byte, error) {
+	endpoint := fmt.Sprintf("http://localhost/containers/%s/json", containerID)
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, endpoint, nil)
+	if err != nil {
+		return nil, err
+	}
+
+	resp, err := dm.client.Do(req)
+	if err != nil {
+		return nil, err
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode != http.StatusOK {
+		body, _ := io.ReadAll(io.LimitReader(resp.Body, 1024))
+		return nil, fmt.Errorf("container info request failed: %s: %s", resp.Status, strings.TrimSpace(string(body)))
+	}
+
+	// Remove sensitive environment variables from Config.Env
+	var containerInfo map[string]any
+	if err := json.NewDecoder(resp.Body).Decode(&containerInfo); err != nil {
+		return nil, err
+	}
+	if config, ok := containerInfo["Config"].(map[string]any); ok {
+		delete(config, "Env")
+	}
+
+	return json.Marshal(containerInfo)
+}
+
+// getLogs fetches the logs for a container
+func (dm *dockerManager) getLogs(ctx context.Context, containerID string) (string, error) {
+	endpoint := fmt.Sprintf("http://localhost/containers/%s/logs?stdout=1&stderr=1&tail=%d", containerID, dockerLogsTail)
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, endpoint, nil)
+	if err != nil {
+		return "", err
+	}
+
+	resp, err := dm.client.Do(req)
+	if err != nil {
+		return "", err
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode != http.StatusOK {
+		body, _ := io.ReadAll(io.LimitReader(resp.Body, 1024))
+		return "", fmt.Errorf("logs request failed: %s: %s", resp.Status, strings.TrimSpace(string(body)))
+	}
+
+	var builder strings.Builder
+	multiplexed := resp.Header.Get("Content-Type") == "application/vnd.docker.multiplexed-stream"
+	if err := decodeDockerLogStream(resp.Body, &builder, multiplexed); err != nil {
+		return "", err
+	}
+
+	// Strip ANSI escape sequences from logs for clean display in web UI
+	logs := builder.String()
+	if strings.Contains(logs, "\x1b") {
+		logs = ansiEscapePattern.ReplaceAllString(logs, "")
+	}
+	return logs, nil
+}
+
+func decodeDockerLogStream(reader io.Reader, builder *strings.Builder, multiplexed bool) error {
+	if !multiplexed {
+		_, err := io.Copy(builder, io.LimitReader(reader, maxTotalLogSize))
+		return err
+	}
+	const headerSize = 8
+	var header [headerSize]byte
+	totalBytesRead := 0
+
+	for {
+		if _, err := io.ReadFull(reader, header[:]); err != nil {
+			if errors.Is(err, io.EOF) || errors.Is(err, io.ErrUnexpectedEOF) {
+				return nil
+			}
+			return err
+		}
+
+		frameLen := binary.BigEndian.Uint32(header[4:])
+		if frameLen == 0 {
+			continue
+		}
+
+		// Prevent memory exhaustion from excessively large frames
+		if frameLen > maxLogFrameSize {
+			return fmt.Errorf("log frame size (%d) exceeds maximum (%d)", frameLen, maxLogFrameSize)
+		}
+
+		// Check if reading this frame would exceed total log size limit
+		if totalBytesRead+int(frameLen) > maxTotalLogSize {
+			// Read and discard remaining data to avoid blocking
+			_, _ = io.CopyN(io.Discard, reader, int64(frameLen))
+			slog.Debug("Truncating logs: limit reached", "read", totalBytesRead, "limit", maxTotalLogSize)
+			return nil
+		}
+
+		n, err := io.CopyN(builder, reader, int64(frameLen))
+		if err != nil {
+			if errors.Is(err, io.EOF) || errors.Is(err, io.ErrUnexpectedEOF) {
+				return nil
+			}
+			return err
+		}
+		totalBytesRead += int(n)
+	}
+}
+
+// GetHostInfo fetches the system info from Docker
+func (dm *dockerManager) GetHostInfo() (info container.HostInfo, err error) {
+	resp, err := dm.client.Get("http://localhost/info")
+	if err != nil {
+		return info, err
+	}
+	defer resp.Body.Close()
+
+	if err := json.NewDecoder(resp.Body).Decode(&info); err != nil {
+		return info, err
+	}
+
+	return info, nil
+}
+
+func (dm *dockerManager) IsPodman() bool {
+	return dm.usingPodman
+}

agent/gpu.go

@@ -1,11 +1,11 @@
 package agent
 
 import (
-	"beszel/internal/entities/system"
 	"bufio"
 	"bytes"
 	"encoding/json"
 	"fmt"
+	"maps"
 	"os/exec"
 	"regexp"
 	"strconv"
@@ -13,7 +13,9 @@ import (
 	"sync"
 	"time"
 
-	"golang.org/x/exp/slog"
+	"github.com/henrygd/beszel/internal/entities/system"
+
+	"log/slog"
 )
 
 const (
@@ -26,13 +28,10 @@ const (
 	nvidiaSmiInterval  string        = "4"    // in seconds
 	tegraStatsInterval string        = "3700" // in milliseconds
 	rocmSmiInterval    time.Duration = 4300 * time.Millisecond
-
 	// Command retry and timeout constants
 	retryWaitTime     time.Duration = 5 * time.Second
 	maxFailureRetries int           = 5
 
-	cmdBufferSize uint16 = 10 * 1024
-
 	// Unit Conversions
 	mebibytesInAMegabyte float64 = 1.024  // nvidia-smi reports memory in MiB
 	milliwattsInAWatt    float64 = 1000.0 // tegrastats reports power in mW
@@ -41,10 +40,27 @@ const (
 // GPUManager manages data collection for GPUs (either Nvidia or AMD)
 type GPUManager struct {
 	sync.Mutex
-	nvidiaSmi  bool
-	rocmSmi    bool
-	tegrastats bool
-	GpuDataMap map[string]*system.GPUData
+	nvidiaSmi     bool
+	rocmSmi       bool
+	tegrastats    bool
+	intelGpuStats bool
+	nvml          bool
+	GpuDataMap    map[string]*system.GPUData
+	// lastAvgData stores the last calculated averages for each GPU
+	// Used when a collection happens before new data arrives (Count == 0)
+	lastAvgData map[string]system.GPUData
+	// Per-cache-key tracking for delta calculations
+	// cacheKey -> gpuId -> snapshot of last count/usage/power values
+	lastSnapshots map[uint16]map[string]*gpuSnapshot
+}
+
+// gpuSnapshot stores the last observed incremental values for delta tracking
+type gpuSnapshot struct {
+	count    uint32
+	usage    float64
+	power    float64
+	powerPkg float64
+	engines  map[string]float64
 }
 
 // RocmSmiJson represents the JSON structure of rocm-smi output
@@ -65,6 +81,7 @@ type gpuCollector struct {
 	cmdArgs []string
 	parse   func([]byte) bool // returns true if valid data was found
 	buf     []byte
+	bufSize uint16
 }
 
 var errNoValidData = fmt.Errorf("no valid GPU data found") // Error for missing data
@@ -98,7 +115,7 @@ func (c *gpuCollector) collect() error {
 
 	scanner := bufio.NewScanner(stdout)
 	if c.buf == nil {
-		c.buf = make([]byte, 0, cmdBufferSize)
+		c.buf = make([]byte, 0, c.bufSize)
 	}
 	scanner.Buffer(c.buf, bufio.MaxScanTokenSize)
 
@@ -229,36 +246,21 @@ func (gm *GPUManager) parseAmdData(output []byte) bool {
 	return true
 }
 
-// sums and resets the current GPU utilization data since the last update
-func (gm *GPUManager) GetCurrentData() map[string]system.GPUData {
+// GetCurrentData returns GPU utilization data averaged since the last call with this cacheKey
+func (gm *GPUManager) GetCurrentData(cacheKey uint16) map[string]system.GPUData {
 	gm.Lock()
 	defer gm.Unlock()
 
-	// check for GPUs with the same name
-	nameCounts := make(map[string]int)
-	for _, gpu := range gm.GpuDataMap {
-		nameCounts[gpu.Name]++
-	}
+	gm.initializeSnapshots(cacheKey)
+	nameCounts := gm.countGPUNames()
 
-	// copy / reset the data
 	gpuData := make(map[string]system.GPUData, len(gm.GpuDataMap))
 	for id, gpu := range gm.GpuDataMap {
-		gpuAvg := *gpu
+		gpuAvg := gm.calculateGPUAverage(id, gpu, cacheKey)
+		gm.updateInstantaneousValues(&gpuAvg, gpu)
+		gm.storeSnapshot(id, gpu, cacheKey)
 
-		gpuAvg.Temperature = twoDecimals(gpu.Temperature)
-		gpuAvg.MemoryUsed = twoDecimals(gpu.MemoryUsed)
-		gpuAvg.MemoryTotal = twoDecimals(gpu.MemoryTotal)
-
-		// avoid division by zero
-		if gpu.Count > 0 {
-			gpuAvg.Usage = twoDecimals(gpu.Usage / gpu.Count)
-			gpuAvg.Power = twoDecimals(gpu.Power / gpu.Count)
-		}
-
-		// reset accumulators in the original
-		gpu.Usage, gpu.Power, gpu.Count = 0, 0, 0
-
-		// append id to the name if there are multiple GPUs with the same name
+		// Append id to name if there are multiple GPUs with the same name
 		if nameCounts[gpu.Name] > 1 {
 			gpuAvg.Name = fmt.Sprintf("%s %s", gpu.Name, id)
 		}
@@ -268,6 +270,120 @@ func (gm *GPUManager) GetCurrentData() map[string]system.GPUData {
 	return gpuData
 }
 
+// initializeSnapshots ensures snapshot maps are initialized for the given cache key
+func (gm *GPUManager) initializeSnapshots(cacheKey uint16) {
+	if gm.lastAvgData == nil {
+		gm.lastAvgData = make(map[string]system.GPUData)
+	}
+	if gm.lastSnapshots == nil {
+		gm.lastSnapshots = make(map[uint16]map[string]*gpuSnapshot)
+	}
+	if gm.lastSnapshots[cacheKey] == nil {
+		gm.lastSnapshots[cacheKey] = make(map[string]*gpuSnapshot)
+	}
+}
+
+// countGPUNames returns a map of GPU names to their occurrence count
+func (gm *GPUManager) countGPUNames() map[string]int {
+	nameCounts := make(map[string]int)
+	for _, gpu := range gm.GpuDataMap {
+		nameCounts[gpu.Name]++
+	}
+	return nameCounts
+}
+
+// calculateGPUAverage computes the average GPU metrics since the last snapshot for this cache key
+func (gm *GPUManager) calculateGPUAverage(id string, gpu *system.GPUData, cacheKey uint16) system.GPUData {
+	lastSnapshot := gm.lastSnapshots[cacheKey][id]
+	currentCount := uint32(gpu.Count)
+	deltaCount := gm.calculateDeltaCount(currentCount, lastSnapshot)
+
+	// If no new data arrived
+	if deltaCount == 0 {
+		// If GPU appears suspended (instantaneous values are 0), return zero values
+		// Otherwise return last known average for temporary collection gaps
+		if gpu.Temperature == 0 && gpu.MemoryUsed == 0 {
+			return system.GPUData{Name: gpu.Name}
+		}
+		return gm.lastAvgData[id] // zero value if not found
+	}
+
+	// Calculate new average
+	gpuAvg := *gpu
+	deltaUsage, deltaPower, deltaPowerPkg := gm.calculateDeltas(gpu, lastSnapshot)
+
+	gpuAvg.Power = twoDecimals(deltaPower / float64(deltaCount))
+
+	if gpu.Engines != nil {
+		// make fresh map for averaged engine metrics to avoid mutating
+		// the accumulator map stored in gm.GpuDataMap
+		gpuAvg.Engines = make(map[string]float64, len(gpu.Engines))
+		gpuAvg.Usage = gm.calculateIntelGPUUsage(&gpuAvg, gpu, lastSnapshot, deltaCount)
+		gpuAvg.PowerPkg = twoDecimals(deltaPowerPkg / float64(deltaCount))
+	} else {
+		gpuAvg.Usage = twoDecimals(deltaUsage / float64(deltaCount))
+	}
+
+	gm.lastAvgData[id] = gpuAvg
+	return gpuAvg
+}
+
+// calculateDeltaCount returns the change in count since the last snapshot
+func (gm *GPUManager) calculateDeltaCount(currentCount uint32, lastSnapshot *gpuSnapshot) uint32 {
+	if lastSnapshot != nil {
+		return currentCount - lastSnapshot.count
+	}
+	return currentCount
+}
+
+// calculateDeltas computes the change in usage, power, and powerPkg since the last snapshot
+func (gm *GPUManager) calculateDeltas(gpu *system.GPUData, lastSnapshot *gpuSnapshot) (deltaUsage, deltaPower, deltaPowerPkg float64) {
+	if lastSnapshot != nil {
+		return gpu.Usage - lastSnapshot.usage,
+			gpu.Power - lastSnapshot.power,
+			gpu.PowerPkg - lastSnapshot.powerPkg
+	}
+	return gpu.Usage, gpu.Power, gpu.PowerPkg
+}
+
+// calculateIntelGPUUsage computes Intel GPU usage from engine metrics and returns max engine usage
+func (gm *GPUManager) calculateIntelGPUUsage(gpuAvg, gpu *system.GPUData, lastSnapshot *gpuSnapshot, deltaCount uint32) float64 {
+	maxEngineUsage := 0.0
+	for name, engine := range gpu.Engines {
+		var deltaEngine float64
+		if lastSnapshot != nil && lastSnapshot.engines != nil {
+			deltaEngine = engine - lastSnapshot.engines[name]
+		} else {
+			deltaEngine = engine
+		}
+		gpuAvg.Engines[name] = twoDecimals(deltaEngine / float64(deltaCount))
+		maxEngineUsage = max(maxEngineUsage, deltaEngine/float64(deltaCount))
+	}
+	return twoDecimals(maxEngineUsage)
+}
+
+// updateInstantaneousValues updates values that should reflect current state, not averages
+func (gm *GPUManager) updateInstantaneousValues(gpuAvg *system.GPUData, gpu *system.GPUData) {
+	gpuAvg.Temperature = twoDecimals(gpu.Temperature)
+	gpuAvg.MemoryUsed = twoDecimals(gpu.MemoryUsed)
+	gpuAvg.MemoryTotal = twoDecimals(gpu.MemoryTotal)
+}
+
+// storeSnapshot saves the current GPU state for this cache key
+func (gm *GPUManager) storeSnapshot(id string, gpu *system.GPUData, cacheKey uint16) {
+	snapshot := &gpuSnapshot{
+		count:    uint32(gpu.Count),
+		usage:    gpu.Usage,
+		power:    gpu.Power,
+		powerPkg: gpu.PowerPkg,
+	}
+	if gpu.Engines != nil {
+		snapshot.engines = make(map[string]float64, len(gpu.Engines))
+		maps.Copy(snapshot.engines, gpu.Engines)
+	}
+	gm.lastSnapshots[cacheKey][id] = snapshot
+}
+
 // detectGPUs checks for the presence of GPU management tools (nvidia-smi, rocm-smi, tegrastats)
 // in the system path. It sets the corresponding flags in the GPUManager struct if any of these
 // tools are found. If none of the tools are found, it returns an error indicating that no GPU
@@ -283,18 +399,37 @@ func (gm *GPUManager) detectGPUs() error {
 		gm.tegrastats = true
 		gm.nvidiaSmi = false
 	}
-	if gm.nvidiaSmi || gm.rocmSmi || gm.tegrastats {
+	if _, err := exec.LookPath(intelGpuStatsCmd); err == nil {
+		gm.intelGpuStats = true
+	}
+	if gm.nvidiaSmi || gm.rocmSmi || gm.tegrastats || gm.intelGpuStats || gm.nvml {
 		return nil
 	}
-	return fmt.Errorf("no GPU found - install nvidia-smi, rocm-smi, or tegrastats")
+	return fmt.Errorf("no GPU found - install nvidia-smi, rocm-smi, tegrastats, or intel_gpu_top")
 }
 
 // startCollector starts the appropriate GPU data collector based on the command
 func (gm *GPUManager) startCollector(command string) {
 	collector := gpuCollector{
-		name: command,
+		name:    command,
+		bufSize: 10 * 1024,
 	}
 	switch command {
+	case intelGpuStatsCmd:
+		go func() {
+			failures := 0
+			for {
+				if err := gm.collectIntelStats(); err != nil {
+					failures++
+					if failures > maxFailureRetries {
+						break
+					}
+					slog.Warn("Error collecting Intel GPU data; see https://beszel.dev/guide/gpu", "err", err)
+					time.Sleep(retryWaitTime)
+					continue
+				}
+			}
+		}()
 	case nvidiaSmiCmd:
 		collector.cmdArgs = []string{
 			"-l", nvidiaSmiInterval,
@@ -328,6 +463,9 @@ func (gm *GPUManager) startCollector(command string) {
 
 // NewGPUManager creates and initializes a new GPUManager
 func NewGPUManager() (*GPUManager, error) {
+	if skipGPU, _ := GetEnv("SKIP_GPU"); skipGPU == "true" {
+		return nil, nil
+	}
 	var gm GPUManager
 	if err := gm.detectGPUs(); err != nil {
 		return nil, err
@@ -335,7 +473,20 @@ func NewGPUManager() (*GPUManager, error) {
 	gm.GpuDataMap = make(map[string]*system.GPUData)
 
 	if gm.nvidiaSmi {
-		gm.startCollector(nvidiaSmiCmd)
+		if nvml, _ := GetEnv("NVML"); nvml == "true" {
+			gm.nvml = true
+			gm.nvidiaSmi = false
+			collector := &nvmlCollector{gm: &gm}
+			if err := collector.init(); err == nil {
+				go collector.start()
+			} else {
+				slog.Warn("Failed to initialize NVML, falling back to nvidia-smi", "err", err)
+				gm.nvidiaSmi = true
+				gm.startCollector(nvidiaSmiCmd)
+			}
+		} else {
+			gm.startCollector(nvidiaSmiCmd)
+		}
 	}
 	if gm.rocmSmi {
 		gm.startCollector(rocmSmiCmd)
@@ -343,6 +494,9 @@ func NewGPUManager() (*GPUManager, error) {
 	if gm.tegrastats {
 		gm.startCollector(tegraStatsCmd)
 	}
+	if gm.intelGpuStats {
+		gm.startCollector(intelGpuStatsCmd)
+	}
 
 	return &gm, nil
 }

agent/gpu_intel.go

@@ -0,0 +1,206 @@
+package agent
+
+import (
+	"bufio"
+	"io"
+	"os/exec"
+	"strconv"
+	"strings"
+
+	"github.com/henrygd/beszel/internal/entities/system"
+)
+
+const (
+	intelGpuStatsCmd      string = "intel_gpu_top"
+	intelGpuStatsInterval string = "3300" // in milliseconds
+)
+
+type intelGpuStats struct {
+	PowerGPU float64
+	PowerPkg float64
+	Engines  map[string]float64
+}
+
+// updateIntelFromStats updates aggregated GPU data from a single intelGpuStats sample
+func (gm *GPUManager) updateIntelFromStats(sample *intelGpuStats) bool {
+	gm.Lock()
+	defer gm.Unlock()
+
+	// only one gpu for now - cmd doesn't provide all by default
+	gpuData, ok := gm.GpuDataMap["0"]
+	if !ok {
+		gpuData = &system.GPUData{Name: "GPU", Engines: make(map[string]float64)}
+		gm.GpuDataMap["0"] = gpuData
+	}
+
+	gpuData.Power += sample.PowerGPU
+	gpuData.PowerPkg += sample.PowerPkg
+
+	if gpuData.Engines == nil {
+		gpuData.Engines = make(map[string]float64, len(sample.Engines))
+	}
+	for name, engine := range sample.Engines {
+		gpuData.Engines[name] += engine
+	}
+
+	gpuData.Count++
+	return true
+}
+
+// collectIntelStats executes intel_gpu_top in text mode (-l) and parses the output
+func (gm *GPUManager) collectIntelStats() (err error) {
+	// Build command arguments, optionally selecting a device via -d
+	args := []string{"-s", intelGpuStatsInterval, "-l"}
+	if dev, ok := GetEnv("INTEL_GPU_DEVICE"); ok && dev != "" {
+		args = append(args, "-d", dev)
+	}
+	cmd := exec.Command(intelGpuStatsCmd, args...)
+	// Avoid blocking if intel_gpu_top writes to stderr
+	cmd.Stderr = io.Discard
+	stdout, err := cmd.StdoutPipe()
+	if err != nil {
+		return err
+	}
+	if err := cmd.Start(); err != nil {
+		return err
+	}
+
+	// Ensure we always reap the child to avoid zombies on any return path and
+	// propagate a non-zero exit code if no other error was set.
+	defer func() {
+		// Best-effort close of the pipe (unblock the child if it writes)
+		_ = stdout.Close()
+		if cmd.ProcessState == nil || !cmd.ProcessState.Exited() {
+			_ = cmd.Process.Kill()
+		}
+		if waitErr := cmd.Wait(); err == nil && waitErr != nil {
+			err = waitErr
+		}
+	}()
+
+	scanner := bufio.NewScanner(stdout)
+	var header1 string
+	var engineNames []string
+	var friendlyNames []string
+	var preEngineCols int
+	var powerIndex int
+	var hadDataRow bool
+	// skip first data row because it sometimes has erroneous data
+	var skippedFirstDataRow bool
+
+	for scanner.Scan() {
+		line := strings.TrimSpace(scanner.Text())
+		if line == "" {
+			continue
+		}
+
+		// first header line
+		if strings.HasPrefix(line, "Freq") {
+			header1 = line
+			continue
+		}
+
+		// second header line
+		if strings.HasPrefix(line, "req") {
+			engineNames, friendlyNames, powerIndex, preEngineCols = gm.parseIntelHeaders(header1, line)
+			continue
+		}
+
+		// Data row
+		if !skippedFirstDataRow {
+			skippedFirstDataRow = true
+			continue
+		}
+		sample, err := gm.parseIntelData(line, engineNames, friendlyNames, powerIndex, preEngineCols)
+		if err != nil {
+			return err
+		}
+		hadDataRow = true
+		gm.updateIntelFromStats(&sample)
+	}
+	if scanErr := scanner.Err(); scanErr != nil {
+		return scanErr
+	}
+	if !hadDataRow {
+		return errNoValidData
+	}
+	return nil
+}
+
+func (gm *GPUManager) parseIntelHeaders(header1 string, header2 string) (engineNames []string, friendlyNames []string, powerIndex int, preEngineCols int) {
+	// Build indexes
+	h1 := strings.Fields(header1)
+	h2 := strings.Fields(header2)
+	powerIndex = -1 // Initialize to -1, will be set to actual index if found
+	// Collect engine names from header1
+	for _, col := range h1 {
+		key := strings.TrimRightFunc(col, func(r rune) bool {
+			return (r >= '0' && r <= '9') || r == '/'
+		})
+		var friendly string
+		switch key {
+		case "RCS":
+			friendly = "Render/3D"
+		case "BCS":
+			friendly = "Blitter"
+		case "VCS":
+			friendly = "Video"
+		case "VECS":
+			friendly = "VideoEnhance"
+		case "CCS":
+			friendly = "Compute"
+		default:
+			continue
+		}
+		engineNames = append(engineNames, key)
+		friendlyNames = append(friendlyNames, friendly)
+	}
+	// find power gpu index among pre-engine columns
+	if n := len(engineNames); n > 0 {
+		preEngineCols = max(len(h2)-3*n, 0)
+		limit := min(len(h2), preEngineCols)
+		for i := range limit {
+			if strings.EqualFold(h2[i], "gpu") {
+				powerIndex = i
+				break
+			}
+		}
+	}
+	return engineNames, friendlyNames, powerIndex, preEngineCols
+}
+
+func (gm *GPUManager) parseIntelData(line string, engineNames []string, friendlyNames []string, powerIndex int, preEngineCols int) (sample intelGpuStats, err error) {
+	fields := strings.Fields(line)
+	if len(fields) == 0 {
+		return sample, errNoValidData
+	}
+	// Make sure row has enough columns for engines
+	if need := preEngineCols + 3*len(engineNames); len(fields) < need {
+		return sample, errNoValidData
+	}
+	if powerIndex >= 0 && powerIndex < len(fields) {
+		if v, perr := strconv.ParseFloat(fields[powerIndex], 64); perr == nil {
+			sample.PowerGPU = v
+		}
+		if v, perr := strconv.ParseFloat(fields[powerIndex+1], 64); perr == nil {
+			sample.PowerPkg = v
+		}
+	}
+	if len(engineNames) > 0 {
+		sample.Engines = make(map[string]float64, len(engineNames))
+		for k := range engineNames {
+			base := preEngineCols + 3*k
+			if base < len(fields) {
+				busy := 0.0
+				if v, e := strconv.ParseFloat(fields[base], 64); e == nil {
+					busy = v
+				}
+				cur := sample.Engines[friendlyNames[k]]
+				sample.Engines[friendlyNames[k]] = cur + busy
+			} else {
+				sample.Engines[friendlyNames[k]] = 0
+			}
+		}
+	}
+	return sample, nil
+}

agent/gpu_nvml.go

@@ -0,0 +1,224 @@
+//go:build (linux || windows) && (amd64 || arm64)
+
+package agent
+
+import (
+	"fmt"
+	"log/slog"
+	"strings"
+	"time"
+	"unsafe"
+
+	"github.com/ebitengine/purego"
+	"github.com/henrygd/beszel/internal/entities/system"
+)
+
+// NVML constants and types
+const (
+	nvmlSuccess int = 0
+)
+
+type nvmlDevice uintptr
+
+type nvmlReturn int
+
+type nvmlMemoryV1 struct {
+	Total uint64
+	Free  uint64
+	Used  uint64
+}
+
+type nvmlMemoryV2 struct {
+	Version  uint32
+	Total    uint64
+	Reserved uint64
+	Free     uint64
+	Used     uint64
+}
+
+type nvmlUtilization struct {
+	Gpu    uint32
+	Memory uint32
+}
+
+type nvmlPciInfo struct {
+	BusId          [16]byte
+	Domain         uint32
+	Bus            uint32
+	Device         uint32
+	PciDeviceId    uint32
+	PciSubSystemId uint32
+}
+
+// NVML function signatures
+var (
+	nvmlInit                      func() nvmlReturn
+	nvmlShutdown                  func() nvmlReturn
+	nvmlDeviceGetCount            func(count *uint32) nvmlReturn
+	nvmlDeviceGetHandleByIndex    func(index uint32, device *nvmlDevice) nvmlReturn
+	nvmlDeviceGetName             func(device nvmlDevice, name *byte, length uint32) nvmlReturn
+	nvmlDeviceGetMemoryInfo       func(device nvmlDevice, memory uintptr) nvmlReturn
+	nvmlDeviceGetUtilizationRates func(device nvmlDevice, utilization *nvmlUtilization) nvmlReturn
+	nvmlDeviceGetTemperature      func(device nvmlDevice, sensorType int, temp *uint32) nvmlReturn
+	nvmlDeviceGetPowerUsage       func(device nvmlDevice, power *uint32) nvmlReturn
+	nvmlDeviceGetPciInfo          func(device nvmlDevice, pci *nvmlPciInfo) nvmlReturn
+	nvmlErrorString               func(result nvmlReturn) string
+)
+
+type nvmlCollector struct {
+	gm      *GPUManager
+	lib     uintptr
+	devices []nvmlDevice
+	bdfs    []string
+	isV2    bool
+}
+
+func (c *nvmlCollector) init() error {
+	slog.Debug("NVML: Initializing")
+	libPath := getNVMLPath()
+
+	lib, err := openLibrary(libPath)
+	if err != nil {
+		return fmt.Errorf("failed to load %s: %w", libPath, err)
+	}
+	c.lib = lib
+
+	purego.RegisterLibFunc(&nvmlInit, lib, "nvmlInit")
+	purego.RegisterLibFunc(&nvmlShutdown, lib, "nvmlShutdown")
+	purego.RegisterLibFunc(&nvmlDeviceGetCount, lib, "nvmlDeviceGetCount")
+	purego.RegisterLibFunc(&nvmlDeviceGetHandleByIndex, lib, "nvmlDeviceGetHandleByIndex")
+	purego.RegisterLibFunc(&nvmlDeviceGetName, lib, "nvmlDeviceGetName")
+	// Try to get v2 memory info, fallback to v1 if not available
+	if hasSymbol(lib, "nvmlDeviceGetMemoryInfo_v2") {
+		c.isV2 = true
+		purego.RegisterLibFunc(&nvmlDeviceGetMemoryInfo, lib, "nvmlDeviceGetMemoryInfo_v2")
+	} else {
+		purego.RegisterLibFunc(&nvmlDeviceGetMemoryInfo, lib, "nvmlDeviceGetMemoryInfo")
+	}
+	purego.RegisterLibFunc(&nvmlDeviceGetUtilizationRates, lib, "nvmlDeviceGetUtilizationRates")
+	purego.RegisterLibFunc(&nvmlDeviceGetTemperature, lib, "nvmlDeviceGetTemperature")
+	purego.RegisterLibFunc(&nvmlDeviceGetPowerUsage, lib, "nvmlDeviceGetPowerUsage")
+	purego.RegisterLibFunc(&nvmlDeviceGetPciInfo, lib, "nvmlDeviceGetPciInfo")
+	purego.RegisterLibFunc(&nvmlErrorString, lib, "nvmlErrorString")
+
+	if ret := nvmlInit(); ret != nvmlReturn(nvmlSuccess) {
+		return fmt.Errorf("nvmlInit failed: %v", ret)
+	}
+
+	var count uint32
+	if ret := nvmlDeviceGetCount(&count); ret != nvmlReturn(nvmlSuccess) {
+		return fmt.Errorf("nvmlDeviceGetCount failed: %v", ret)
+	}
+
+	for i := uint32(0); i < count; i++ {
+		var device nvmlDevice
+		if ret := nvmlDeviceGetHandleByIndex(i, &device); ret == nvmlReturn(nvmlSuccess) {
+			c.devices = append(c.devices, device)
+			// Get BDF for power state check
+			var pci nvmlPciInfo
+			if ret := nvmlDeviceGetPciInfo(device, &pci); ret == nvmlReturn(nvmlSuccess) {
+				busID := string(pci.BusId[:])
+				if idx := strings.Index(busID, "\x00"); idx != -1 {
+					busID = busID[:idx]
+				}
+				c.bdfs = append(c.bdfs, strings.ToLower(busID))
+			} else {
+				c.bdfs = append(c.bdfs, "")
+			}
+		}
+	}
+
+	return nil
+}
+
+func (c *nvmlCollector) start() {
+	defer nvmlShutdown()
+	ticker := time.Tick(3 * time.Second)
+
+	for range ticker {
+		c.collect()
+	}
+}
+
+func (c *nvmlCollector) collect() {
+	c.gm.Lock()
+	defer c.gm.Unlock()
+
+	for i, device := range c.devices {
+		id := fmt.Sprintf("%d", i)
+		bdf := c.bdfs[i]
+
+		// Update GPUDataMap
+		if _, ok := c.gm.GpuDataMap[id]; !ok {
+			var nameBuf [64]byte
+			if ret := nvmlDeviceGetName(device, &nameBuf[0], 64); ret != nvmlReturn(nvmlSuccess) {
+				continue
+			}
+			name := string(nameBuf[:strings.Index(string(nameBuf[:]), "\x00")])
+			name = strings.TrimPrefix(name, "NVIDIA ")
+			c.gm.GpuDataMap[id] = &system.GPUData{Name: strings.TrimSuffix(name, " Laptop GPU")}
+		}
+		gpu := c.gm.GpuDataMap[id]
+
+		if bdf != "" && !c.isGPUActive(bdf) {
+			slog.Debug("NVML: GPU is suspended, skipping", "bdf", bdf)
+			gpu.Temperature = 0
+			gpu.MemoryUsed = 0
+			continue
+		}
+
+		// Utilization
+		var utilization nvmlUtilization
+		if ret := nvmlDeviceGetUtilizationRates(device, &utilization); ret != nvmlReturn(nvmlSuccess) {
+			slog.Debug("NVML: Utilization failed (GPU likely suspended)", "bdf", bdf, "ret", ret)
+			gpu.Temperature = 0
+			gpu.MemoryUsed = 0
+			continue
+		}
+
+		slog.Debug("NVML: Collecting data for GPU", "bdf", bdf)
+
+		// Temperature
+		var temp uint32
+		nvmlDeviceGetTemperature(device, 0, &temp) // 0 is NVML_TEMPERATURE_GPU
+
+		// Memory: only poll if GPU is active to avoid leaving D3cold state (#1522)
+		if utilization.Gpu > 0 {
+			var usedMem, totalMem uint64
+			if c.isV2 {
+				var memory nvmlMemoryV2
+				memory.Version = 0x02000028 // (2 << 24) | 40 bytes
+				if ret := nvmlDeviceGetMemoryInfo(device, uintptr(unsafe.Pointer(&memory))); ret != nvmlReturn(nvmlSuccess) {
+					slog.Debug("NVML: MemoryInfo_v2 failed", "bdf", bdf, "ret", ret)
+				} else {
+					usedMem = memory.Used
+					totalMem = memory.Total
+				}
+			} else {
+				var memory nvmlMemoryV1
+				if ret := nvmlDeviceGetMemoryInfo(device, uintptr(unsafe.Pointer(&memory))); ret != nvmlReturn(nvmlSuccess) {
+					slog.Debug("NVML: MemoryInfo failed", "bdf", bdf, "ret", ret)
+				} else {
+					usedMem = memory.Used
+					totalMem = memory.Total
+				}
+			}
+			if totalMem > 0 {
+				gpu.MemoryUsed = float64(usedMem) / 1024 / 1024 / mebibytesInAMegabyte
+				gpu.MemoryTotal = float64(totalMem) / 1024 / 1024 / mebibytesInAMegabyte
+			}
+		} else {
+			slog.Debug("NVML: Skipping memory info (utilization=0)", "bdf", bdf)
+		}
+
+		// Power
+		var power uint32
+		nvmlDeviceGetPowerUsage(device, &power)
+
+		gpu.Temperature = float64(temp)
+		gpu.Usage += float64(utilization.Gpu)
+		gpu.Power += float64(power) / 1000.0
+		gpu.Count++
+		slog.Debug("NVML: Collected data", "gpu", gpu)
+	}
+}

agent/gpu_nvml_linux.go

@@ -0,0 +1,57 @@
+//go:build linux && (amd64 || arm64)
+
+package agent
+
+import (
+	"log/slog"
+	"os"
+	"path/filepath"
+	"strings"
+
+	"github.com/ebitengine/purego"
+)
+
+func openLibrary(name string) (uintptr, error) {
+	return purego.Dlopen(name, purego.RTLD_NOW|purego.RTLD_GLOBAL)
+}
+
+func getNVMLPath() string {
+	return "libnvidia-ml.so.1"
+}
+
+func hasSymbol(lib uintptr, symbol string) bool {
+	_, err := purego.Dlsym(lib, symbol)
+	return err == nil
+}
+
+func (c *nvmlCollector) isGPUActive(bdf string) bool {
+	// runtime_status
+	statusPath := filepath.Join("/sys/bus/pci/devices", bdf, "power/runtime_status")
+	status, err := os.ReadFile(statusPath)
+	if err != nil {
+		slog.Debug("NVML: Can't read runtime_status", "bdf", bdf, "err", err)
+		return true // Assume active if we can't read status
+	}
+	statusStr := strings.TrimSpace(string(status))
+	if statusStr != "active" && statusStr != "resuming" {
+		slog.Debug("NVML: GPU not active", "bdf", bdf, "status", statusStr)
+		return false
+	}
+
+	// power_state (D0 check)
+	// Find any drm card device power_state
+	pstatePathPattern := filepath.Join("/sys/bus/pci/devices", bdf, "drm/card*/device/power_state")
+	matches, _ := filepath.Glob(pstatePathPattern)
+	if len(matches) > 0 {
+		pstate, err := os.ReadFile(matches[0])
+		if err == nil {
+			pstateStr := strings.TrimSpace(string(pstate))
+			if pstateStr != "D0" {
+				slog.Debug("NVML: GPU not in D0 state", "bdf", bdf, "pstate", pstateStr)
+				return false
+			}
+		}
+	}
+
+	return true
+}

agent/gpu_nvml_unsupported.go

@@ -0,0 +1,33 @@
+//go:build (!linux && !windows) || (!amd64 && !arm64)
+
+package agent
+
+import "fmt"
+
+type nvmlCollector struct {
+	gm *GPUManager
+}
+
+func (c *nvmlCollector) init() error {
+	return fmt.Errorf("nvml not supported on this platform")
+}
+
+func (c *nvmlCollector) start() {}
+
+func (c *nvmlCollector) collect() {}
+
+func openLibrary(name string) (uintptr, error) {
+	return 0, fmt.Errorf("nvml not supported on this platform")
+}
+
+func getNVMLPath() string {
+	return ""
+}
+
+func hasSymbol(lib uintptr, symbol string) bool {
+	return false
+}
+
+func (c *nvmlCollector) isGPUActive(bdf string) bool {
+	return true
+}

agent/gpu_nvml_windows.go

@@ -0,0 +1,25 @@
+//go:build windows && (amd64 || arm64)
+
+package agent
+
+import (
+	"golang.org/x/sys/windows"
+)
+
+func openLibrary(name string) (uintptr, error) {
+	handle, err := windows.LoadLibrary(name)
+	return uintptr(handle), err
+}
+
+func getNVMLPath() string {
+	return "nvml.dll"
+}
+
+func hasSymbol(lib uintptr, symbol string) bool {
+	_, err := windows.GetProcAddress(windows.Handle(lib), symbol)
+	return err == nil
+}
+
+func (c *nvmlCollector) isGPUActive(bdf string) bool {
+	return true
+}

agent/handlers.go

@@ -0,0 +1,205 @@
+package agent
+
+import (
+	"context"
+	"errors"
+	"fmt"
+
+	"github.com/fxamacker/cbor/v2"
+	"github.com/henrygd/beszel/internal/common"
+	"github.com/henrygd/beszel/internal/entities/smart"
+
+	"log/slog"
+)
+
+// HandlerContext provides context for request handlers
+type HandlerContext struct {
+	Client      *WebSocketClient
+	Agent       *Agent
+	Request     *common.HubRequest[cbor.RawMessage]
+	RequestID   *uint32
+	HubVerified bool
+	// SendResponse abstracts how a handler sends responses (WS or SSH)
+	SendResponse func(data any, requestID *uint32) error
+}
+
+// RequestHandler defines the interface for handling specific websocket request types
+type RequestHandler interface {
+	// Handle processes the request and returns an error if unsuccessful
+	Handle(hctx *HandlerContext) error
+}
+
+// Responder sends handler responses back to the hub (over WS or SSH)
+type Responder interface {
+	SendResponse(data any, requestID *uint32) error
+}
+
+// HandlerRegistry manages the mapping between actions and their handlers
+type HandlerRegistry struct {
+	handlers map[common.WebSocketAction]RequestHandler
+}
+
+// NewHandlerRegistry creates a new handler registry with default handlers
+func NewHandlerRegistry() *HandlerRegistry {
+	registry := &HandlerRegistry{
+		handlers: make(map[common.WebSocketAction]RequestHandler),
+	}
+
+	registry.Register(common.GetData, &GetDataHandler{})
+	registry.Register(common.CheckFingerprint, &CheckFingerprintHandler{})
+	registry.Register(common.GetContainerLogs, &GetContainerLogsHandler{})
+	registry.Register(common.GetContainerInfo, &GetContainerInfoHandler{})
+	registry.Register(common.GetSmartData, &GetSmartDataHandler{})
+	registry.Register(common.GetSystemdInfo, &GetSystemdInfoHandler{})
+
+	return registry
+}
+
+// Register registers a handler for a specific action type
+func (hr *HandlerRegistry) Register(action common.WebSocketAction, handler RequestHandler) {
+	hr.handlers[action] = handler
+}
+
+// Handle routes the request to the appropriate handler
+func (hr *HandlerRegistry) Handle(hctx *HandlerContext) error {
+	handler, exists := hr.handlers[hctx.Request.Action]
+	if !exists {
+		return fmt.Errorf("unknown action: %d", hctx.Request.Action)
+	}
+
+	// Check verification requirement - default to requiring verification
+	if hctx.Request.Action != common.CheckFingerprint && !hctx.HubVerified {
+		return errors.New("hub not verified")
+	}
+
+	// Log handler execution for debugging
+	// slog.Debug("Executing handler", "action", hctx.Request.Action)
+
+	return handler.Handle(hctx)
+}
+
+// GetHandler returns the handler for a specific action
+func (hr *HandlerRegistry) GetHandler(action common.WebSocketAction) (RequestHandler, bool) {
+	handler, exists := hr.handlers[action]
+	return handler, exists
+}
+
+////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////
+
+// GetDataHandler handles system data requests
+type GetDataHandler struct{}
+
+func (h *GetDataHandler) Handle(hctx *HandlerContext) error {
+	var options common.DataRequestOptions
+	_ = cbor.Unmarshal(hctx.Request.Data, &options)
+
+	sysStats := hctx.Agent.gatherStats(options)
+	return hctx.SendResponse(sysStats, hctx.RequestID)
+}
+
+////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////
+
+// CheckFingerprintHandler handles authentication challenges
+type CheckFingerprintHandler struct{}
+
+func (h *CheckFingerprintHandler) Handle(hctx *HandlerContext) error {
+	return hctx.Client.handleAuthChallenge(hctx.Request, hctx.RequestID)
+}
+
+////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////
+
+// GetContainerLogsHandler handles container log requests
+type GetContainerLogsHandler struct{}
+
+func (h *GetContainerLogsHandler) Handle(hctx *HandlerContext) error {
+	if hctx.Agent.dockerManager == nil {
+		return hctx.SendResponse("", hctx.RequestID)
+	}
+
+	var req common.ContainerLogsRequest
+	if err := cbor.Unmarshal(hctx.Request.Data, &req); err != nil {
+		return err
+	}
+
+	ctx := context.Background()
+	logContent, err := hctx.Agent.dockerManager.getLogs(ctx, req.ContainerID)
+	if err != nil {
+		return err
+	}
+
+	return hctx.SendResponse(logContent, hctx.RequestID)
+}
+
+////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////
+
+// GetContainerInfoHandler handles container info requests
+type GetContainerInfoHandler struct{}
+
+func (h *GetContainerInfoHandler) Handle(hctx *HandlerContext) error {
+	if hctx.Agent.dockerManager == nil {
+		return hctx.SendResponse("", hctx.RequestID)
+	}
+
+	var req common.ContainerInfoRequest
+	if err := cbor.Unmarshal(hctx.Request.Data, &req); err != nil {
+		return err
+	}
+
+	ctx := context.Background()
+	info, err := hctx.Agent.dockerManager.getContainerInfo(ctx, req.ContainerID)
+	if err != nil {
+		return err
+	}
+
+	return hctx.SendResponse(string(info), hctx.RequestID)
+}
+
+////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////
+
+// GetSmartDataHandler handles SMART data requests
+type GetSmartDataHandler struct{}
+
+func (h *GetSmartDataHandler) Handle(hctx *HandlerContext) error {
+	if hctx.Agent.smartManager == nil {
+		// return empty map to indicate no data
+		return hctx.SendResponse(map[string]smart.SmartData{}, hctx.RequestID)
+	}
+	if err := hctx.Agent.smartManager.Refresh(false); err != nil {
+		slog.Debug("smart refresh failed", "err", err)
+	}
+	data := hctx.Agent.smartManager.GetCurrentData()
+	return hctx.SendResponse(data, hctx.RequestID)
+}
+
+////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////
+////////////////////////////////////////////////////////////////////////////
+
+// GetSystemdInfoHandler handles detailed systemd service info requests
+type GetSystemdInfoHandler struct{}
+
+func (h *GetSystemdInfoHandler) Handle(hctx *HandlerContext) error {
+	if hctx.Agent.systemdManager == nil {
+		return errors.ErrUnsupported
+	}
+
+	var req common.SystemdInfoRequest
+	if err := cbor.Unmarshal(hctx.Request.Data, &req); err != nil {
+		return err
+	}
+	if req.ServiceName == "" {
+		return errors.New("service name is required")
+	}
+
+	details, err := hctx.Agent.systemdManager.getServiceDetails(req.ServiceName)
+	if err != nil {
+		return err
+	}
+
+	return hctx.SendResponse(details, hctx.RequestID)
+}

agent/handlers_test.go

@@ -0,0 +1,112 @@
+//go:build testing
+// +build testing
+
+package agent
+
+import (
+	"testing"
+
+	"github.com/fxamacker/cbor/v2"
+	"github.com/henrygd/beszel/internal/common"
+	"github.com/stretchr/testify/assert"
+)
+
+// MockHandler for testing
+type MockHandler struct {
+	requiresVerification bool
+	description          string
+	handleFunc           func(ctx *HandlerContext) error
+}
+
+func (m *MockHandler) Handle(ctx *HandlerContext) error {
+	if m.handleFunc != nil {
+		return m.handleFunc(ctx)
+	}
+	return nil
+}
+
+func (m *MockHandler) RequiresVerification() bool {
+	return m.requiresVerification
+}
+
+// TestHandlerRegistry tests the handler registry functionality
+func TestHandlerRegistry(t *testing.T) {
+	t.Run("default registration", func(t *testing.T) {
+		registry := NewHandlerRegistry()
+
+		// Check default handlers are registered
+		getDataHandler, exists := registry.GetHandler(common.GetData)
+		assert.True(t, exists)
+		assert.IsType(t, &GetDataHandler{}, getDataHandler)
+
+		fingerprintHandler, exists := registry.GetHandler(common.CheckFingerprint)
+		assert.True(t, exists)
+		assert.IsType(t, &CheckFingerprintHandler{}, fingerprintHandler)
+	})
+
+	t.Run("custom handler registration", func(t *testing.T) {
+		registry := NewHandlerRegistry()
+		mockHandler := &MockHandler{
+			requiresVerification: true,
+			description:          "Test handler",
+		}
+
+		// Register a custom handler for a mock action
+		const mockAction common.WebSocketAction = 99
+		registry.Register(mockAction, mockHandler)
+
+		// Verify registration
+		handler, exists := registry.GetHandler(mockAction)
+		assert.True(t, exists)
+		assert.Equal(t, mockHandler, handler)
+	})
+
+	t.Run("unknown action", func(t *testing.T) {
+		registry := NewHandlerRegistry()
+		ctx := &HandlerContext{
+			Request: &common.HubRequest[cbor.RawMessage]{
+				Action: common.WebSocketAction(255), // Unknown action
+			},
+			HubVerified: true,
+		}
+
+		err := registry.Handle(ctx)
+		assert.Error(t, err)
+		assert.Contains(t, err.Error(), "unknown action: 255")
+	})
+
+	t.Run("verification required", func(t *testing.T) {
+		registry := NewHandlerRegistry()
+		ctx := &HandlerContext{
+			Request: &common.HubRequest[cbor.RawMessage]{
+				Action: common.GetData, // Requires verification
+			},
+			HubVerified: false, // Not verified
+		}
+
+		err := registry.Handle(ctx)
+		assert.Error(t, err)
+		assert.Contains(t, err.Error(), "hub not verified")
+	})
+}
+
+// TestCheckFingerprintHandler tests the CheckFingerprint handler
+func TestCheckFingerprintHandler(t *testing.T) {
+	handler := &CheckFingerprintHandler{}
+
+	t.Run("handle with invalid data", func(t *testing.T) {
+		client := &WebSocketClient{}
+		ctx := &HandlerContext{
+			Client:      client,
+			HubVerified: false,
+			Request: &common.HubRequest[cbor.RawMessage]{
+				Action: common.CheckFingerprint,
+				Data:   cbor.RawMessage{}, // Empty/invalid data
+			},
+		}
+
+		// Should fail to decode the fingerprint request
+		err := handler.Handle(ctx)
+		assert.Error(t, err)
+	})
+}

agent/network.go

@@ -0,0 +1,259 @@
+package agent
+
+import (
+	"fmt"
+	"log/slog"
+	"path"
+	"strings"
+	"time"
+
+	"github.com/henrygd/beszel/agent/deltatracker"
+	"github.com/henrygd/beszel/internal/entities/system"
+	psutilNet "github.com/shirou/gopsutil/v4/net"
+)
+
+// NicConfig controls inclusion/exclusion of network interfaces via the NICS env var
+//
+// Behavior mirrors SensorConfig's matching logic:
+// - Leading '-' means blacklist mode; otherwise whitelist mode
+// - Supports '*' wildcards using path.Match
+// - In whitelist mode with an empty list, no NICs are selected
+// - In blacklist mode with an empty list, all NICs are selected
+type NicConfig struct {
+	nics         map[string]struct{}
+	isBlacklist  bool
+	hasWildcards bool
+}
+
+func newNicConfig(nicsEnvVal string) *NicConfig {
+	cfg := &NicConfig{
+		nics: make(map[string]struct{}),
+	}
+	if strings.HasPrefix(nicsEnvVal, "-") {
+		cfg.isBlacklist = true
+		nicsEnvVal = nicsEnvVal[1:]
+	}
+	for nic := range strings.SplitSeq(nicsEnvVal, ",") {
+		nic = strings.TrimSpace(nic)
+		if nic != "" {
+			cfg.nics[nic] = struct{}{}
+			if strings.Contains(nic, "*") {
+				cfg.hasWildcards = true
+			}
+		}
+	}
+	return cfg
+}
+
+// isValidNic determines if a NIC should be included based on NicConfig rules
+func isValidNic(nicName string, cfg *NicConfig) bool {
+	// Empty list behavior differs by mode: blacklist: allow all; whitelist: allow none
+	if len(cfg.nics) == 0 {
+		return cfg.isBlacklist
+	}
+
+	// Exact match: return true if whitelist, false if blacklist
+	if _, exactMatch := cfg.nics[nicName]; exactMatch {
+		return !cfg.isBlacklist
+	}
+
+	// If no wildcards, return true if blacklist, false if whitelist
+	if !cfg.hasWildcards {
+		return cfg.isBlacklist
+	}
+
+	// Check for wildcard patterns
+	for pattern := range cfg.nics {
+		if !strings.Contains(pattern, "*") {
+			continue
+		}
+		if match, _ := path.Match(pattern, nicName); match {
+			return !cfg.isBlacklist
+		}
+	}
+
+	return cfg.isBlacklist
+}
+
+func (a *Agent) updateNetworkStats(cacheTimeMs uint16, systemStats *system.Stats) {
+	// network stats
+	a.ensureNetInterfacesInitialized()
+
+	a.ensureNetworkInterfacesMap(systemStats)
+
+	if netIO, err := psutilNet.IOCounters(true); err == nil {
+		nis, msElapsed := a.loadAndTickNetBaseline(cacheTimeMs)
+		totalBytesSent, totalBytesRecv := a.sumAndTrackPerNicDeltas(cacheTimeMs, msElapsed, netIO, systemStats)
+		bytesSentPerSecond, bytesRecvPerSecond := a.computeBytesPerSecond(msElapsed, totalBytesSent, totalBytesRecv, nis)
+		a.applyNetworkTotals(cacheTimeMs, netIO, systemStats, nis, totalBytesSent, totalBytesRecv, bytesSentPerSecond, bytesRecvPerSecond)
+	}
+}
+
+func (a *Agent) initializeNetIoStats() {
+	// reset valid network interfaces
+	a.netInterfaces = make(map[string]struct{}, 0)
+
+	// parse NICS env var for whitelist / blacklist
+	nicsEnvVal, nicsEnvExists := GetEnv("NICS")
+	var nicCfg *NicConfig
+	if nicsEnvExists {
+		nicCfg = newNicConfig(nicsEnvVal)
+	}
+
+	// get current network I/O stats and record valid interfaces
+	if netIO, err := psutilNet.IOCounters(true); err == nil {
+		for _, v := range netIO {
+			if nicsEnvExists && !isValidNic(v.Name, nicCfg) {
+				continue
+			}
+			if a.skipNetworkInterface(v) {
+				continue
+			}
+			slog.Info("Detected network interface", "name", v.Name, "sent", v.BytesSent, "recv", v.BytesRecv)
+			// store as a valid network interface
+			a.netInterfaces[v.Name] = struct{}{}
+		}
+	}
+
+	// Reset per-cache-time trackers and baselines so they will reinitialize on next use
+	a.netInterfaceDeltaTrackers = make(map[uint16]*deltatracker.DeltaTracker[string, uint64])
+	a.netIoStats = make(map[uint16]system.NetIoStats)
+}
+
+// ensureNetInterfacesInitialized re-initializes NICs if none are currently tracked
+func (a *Agent) ensureNetInterfacesInitialized() {
+	if len(a.netInterfaces) == 0 {
+		// if no network interfaces, initialize again
+		// this is a fix if agent started before network is online (#466)
+		// maybe refactor this in the future to not cache interface names at all so we
+		// don't miss an interface that's been added after agent started in any circumstance
+		a.initializeNetIoStats()
+	}
+}
+
+// ensureNetworkInterfacesMap ensures systemStats.NetworkInterfaces map exists
+func (a *Agent) ensureNetworkInterfacesMap(systemStats *system.Stats) {
+	if systemStats.NetworkInterfaces == nil {
+		systemStats.NetworkInterfaces = make(map[string][4]uint64, 0)
+	}
+}
+
+// loadAndTickNetBaseline returns the NetIoStats baseline and milliseconds elapsed, updating time
+func (a *Agent) loadAndTickNetBaseline(cacheTimeMs uint16) (netIoStat system.NetIoStats, msElapsed uint64) {
+	netIoStat = a.netIoStats[cacheTimeMs]
+	if netIoStat.Time.IsZero() {
+		netIoStat.Time = time.Now()
+		msElapsed = 0
+	} else {
+		msElapsed = uint64(time.Since(netIoStat.Time).Milliseconds())
+		netIoStat.Time = time.Now()
+	}
+	return netIoStat, msElapsed
+}
+
+// sumAndTrackPerNicDeltas accumulates totals and records per-NIC up/down deltas into systemStats
+func (a *Agent) sumAndTrackPerNicDeltas(cacheTimeMs uint16, msElapsed uint64, netIO []psutilNet.IOCountersStat, systemStats *system.Stats) (totalBytesSent, totalBytesRecv uint64) {
+	tracker := a.netInterfaceDeltaTrackers[cacheTimeMs]
+	if tracker == nil {
+		tracker = deltatracker.NewDeltaTracker[string, uint64]()
+		a.netInterfaceDeltaTrackers[cacheTimeMs] = tracker
+	}
+	tracker.Cycle()
+
+	for _, v := range netIO {
+		if _, exists := a.netInterfaces[v.Name]; !exists {
+			continue
+		}
+		totalBytesSent += v.BytesSent
+		totalBytesRecv += v.BytesRecv
+
+		var upDelta, downDelta uint64
+		upKey, downKey := fmt.Sprintf("%sup", v.Name), fmt.Sprintf("%sdown", v.Name)
+		tracker.Set(upKey, v.BytesSent)
+		tracker.Set(downKey, v.BytesRecv)
+		if msElapsed > 0 {
+			if prevVal, ok := tracker.Previous(upKey); ok {
+				var deltaBytes uint64
+				if v.BytesSent >= prevVal {
+					deltaBytes = v.BytesSent - prevVal
+				} else {
+					deltaBytes = v.BytesSent
+				}
+				upDelta = deltaBytes * 1000 / msElapsed
+			}
+			if prevVal, ok := tracker.Previous(downKey); ok {
+				var deltaBytes uint64
+				if v.BytesRecv >= prevVal {
+					deltaBytes = v.BytesRecv - prevVal
+				} else {
+					deltaBytes = v.BytesRecv
+				}
+				downDelta = deltaBytes * 1000 / msElapsed
+			}
+		}
+		systemStats.NetworkInterfaces[v.Name] = [4]uint64{upDelta, downDelta, v.BytesSent, v.BytesRecv}
+	}
+
+	return totalBytesSent, totalBytesRecv
+}
+
+// computeBytesPerSecond calculates per-second totals from elapsed time and totals
+func (a *Agent) computeBytesPerSecond(msElapsed, totalBytesSent, totalBytesRecv uint64, nis system.NetIoStats) (bytesSentPerSecond, bytesRecvPerSecond uint64) {
+	if msElapsed > 0 {
+		bytesSentPerSecond = (totalBytesSent - nis.BytesSent) * 1000 / msElapsed
+		bytesRecvPerSecond = (totalBytesRecv - nis.BytesRecv) * 1000 / msElapsed
+	}
+	return bytesSentPerSecond, bytesRecvPerSecond
+}
+
+// applyNetworkTotals validates and writes computed network stats, or resets on anomaly
+func (a *Agent) applyNetworkTotals(
+	cacheTimeMs uint16,
+	netIO []psutilNet.IOCountersStat,
+	systemStats *system.Stats,
+	nis system.NetIoStats,
+	totalBytesSent, totalBytesRecv uint64,
+	bytesSentPerSecond, bytesRecvPerSecond uint64,
+) {
+	networkSentPs := bytesToMegabytes(float64(bytesSentPerSecond))
+	networkRecvPs := bytesToMegabytes(float64(bytesRecvPerSecond))
+	if networkSentPs > 10_000 || networkRecvPs > 10_000 {
+		slog.Warn("Invalid net stats. Resetting.", "sent", networkSentPs, "recv", networkRecvPs)
+		for _, v := range netIO {
+			if _, exists := a.netInterfaces[v.Name]; !exists {
+				continue
+			}
+			slog.Info(v.Name, "recv", v.BytesRecv, "sent", v.BytesSent)
+		}
+		a.initializeNetIoStats()
+		delete(a.netIoStats, cacheTimeMs)
+		delete(a.netInterfaceDeltaTrackers, cacheTimeMs)
+		systemStats.NetworkSent = 0
+		systemStats.NetworkRecv = 0
+		systemStats.Bandwidth[0], systemStats.Bandwidth[1] = 0, 0
+		return
+	}
+
+	systemStats.NetworkSent = networkSentPs
+	systemStats.NetworkRecv = networkRecvPs
+	systemStats.Bandwidth[0], systemStats.Bandwidth[1] = bytesSentPerSecond, bytesRecvPerSecond
+	nis.BytesSent = totalBytesSent
+	nis.BytesRecv = totalBytesRecv
+	a.netIoStats[cacheTimeMs] = nis
+}
+
+func (a *Agent) skipNetworkInterface(v psutilNet.IOCountersStat) bool {
+	switch {
+	case strings.HasPrefix(v.Name, "lo"),
+		strings.HasPrefix(v.Name, "docker"),
+		strings.HasPrefix(v.Name, "br-"),
+		strings.HasPrefix(v.Name, "veth"),
+		strings.HasPrefix(v.Name, "bond"),
+		strings.HasPrefix(v.Name, "cali"),
+		v.BytesRecv == 0,
+		v.BytesSent == 0:
+		return true
+	default:
+		return false
+	}
+}

agent/network_test.go

@@ -0,0 +1,502 @@
+//go:build testing
+
+package agent
+
+import (
+	"testing"
+	"time"
+
+	"github.com/henrygd/beszel/agent/deltatracker"
+	"github.com/henrygd/beszel/internal/entities/system"
+	psutilNet "github.com/shirou/gopsutil/v4/net"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestIsValidNic(t *testing.T) {
+	tests := []struct {
+		name          string
+		nicName       string
+		config        *NicConfig
+		expectedValid bool
+	}{
+		{
+			name:    "Whitelist - NIC in list",
+			nicName: "eth0",
+			config: &NicConfig{
+				nics:        map[string]struct{}{"eth0": {}},
+				isBlacklist: false,
+			},
+			expectedValid: true,
+		},
+		{
+			name:    "Whitelist - NIC not in list",
+			nicName: "wlan0",
+			config: &NicConfig{
+				nics:        map[string]struct{}{"eth0": {}},
+				isBlacklist: false,
+			},
+			expectedValid: false,
+		},
+		{
+			name:    "Blacklist - NIC in list",
+			nicName: "eth0",
+			config: &NicConfig{
+				nics:        map[string]struct{}{"eth0": {}},
+				isBlacklist: true,
+			},
+			expectedValid: false,
+		},
+		{
+			name:    "Blacklist - NIC not in list",
+			nicName: "wlan0",
+			config: &NicConfig{
+				nics:        map[string]struct{}{"eth0": {}},
+				isBlacklist: true,
+			},
+			expectedValid: true,
+		},
+		{
+			name:    "Whitelist with wildcard - matching pattern",
+			nicName: "eth1",
+			config: &NicConfig{
+				nics:         map[string]struct{}{"eth*": {}},
+				isBlacklist:  false,
+				hasWildcards: true,
+			},
+			expectedValid: true,
+		},
+		{
+			name:    "Whitelist with wildcard - non-matching pattern",
+			nicName: "wlan0",
+			config: &NicConfig{
+				nics:         map[string]struct{}{"eth*": {}},
+				isBlacklist:  false,
+				hasWildcards: true,
+			},
+			expectedValid: false,
+		},
+		{
+			name:    "Blacklist with wildcard - matching pattern",
+			nicName: "eth1",
+			config: &NicConfig{
+				nics:         map[string]struct{}{"eth*": {}},
+				isBlacklist:  true,
+				hasWildcards: true,
+			},
+			expectedValid: false,
+		},
+		{
+			name:    "Blacklist with wildcard - non-matching pattern",
+			nicName: "wlan0",
+			config: &NicConfig{
+				nics:         map[string]struct{}{"eth*": {}},
+				isBlacklist:  true,
+				hasWildcards: true,
+			},
+			expectedValid: true,
+		},
+		{
+			name:    "Empty whitelist config - no NICs allowed",
+			nicName: "eth0",
+			config: &NicConfig{
+				nics:        map[string]struct{}{},
+				isBlacklist: false,
+			},
+			expectedValid: false,
+		},
+		{
+			name:    "Empty blacklist config - all NICs allowed",
+			nicName: "eth0",
+			config: &NicConfig{
+				nics:        map[string]struct{}{},
+				isBlacklist: true,
+			},
+			expectedValid: true,
+		},
+		{
+			name:    "Multiple patterns - exact match",
+			nicName: "eth0",
+			config: &NicConfig{
+				nics:        map[string]struct{}{"eth0": {}, "wlan*": {}},
+				isBlacklist: false,
+			},
+			expectedValid: true,
+		},
+		{
+			name:    "Multiple patterns - wildcard match",
+			nicName: "wlan1",
+			config: &NicConfig{
+				nics:         map[string]struct{}{"eth0": {}, "wlan*": {}},
+				isBlacklist:  false,
+				hasWildcards: true,
+			},
+			expectedValid: true,
+		},
+		{
+			name:    "Multiple patterns - no match",
+			nicName: "bond0",
+			config: &NicConfig{
+				nics:         map[string]struct{}{"eth0": {}, "wlan*": {}},
+				isBlacklist:  false,
+				hasWildcards: true,
+			},
+			expectedValid: false,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			result := isValidNic(tt.nicName, tt.config)
+			assert.Equal(t, tt.expectedValid, result)
+		})
+	}
+}
+
+func TestNewNicConfig(t *testing.T) {
+	tests := []struct {
+		name        string
+		nicsEnvVal  string
+		expectedCfg *NicConfig
+	}{
+		{
+			name:       "Empty string",
+			nicsEnvVal: "",
+			expectedCfg: &NicConfig{
+				nics:         map[string]struct{}{},
+				isBlacklist:  false,
+				hasWildcards: false,
+			},
+		},
+		{
+			name:       "Single NIC whitelist",
+			nicsEnvVal: "eth0",
+			expectedCfg: &NicConfig{
+				nics:         map[string]struct{}{"eth0": {}},
+				isBlacklist:  false,
+				hasWildcards: false,
+			},
+		},
+		{
+			name:       "Multiple NICs whitelist",
+			nicsEnvVal: "eth0,wlan0",
+			expectedCfg: &NicConfig{
+				nics:         map[string]struct{}{"eth0": {}, "wlan0": {}},
+				isBlacklist:  false,
+				hasWildcards: false,
+			},
+		},
+		{
+			name:       "Blacklist mode",
+			nicsEnvVal: "-eth0,wlan0",
+			expectedCfg: &NicConfig{
+				nics:         map[string]struct{}{"eth0": {}, "wlan0": {}},
+				isBlacklist:  true,
+				hasWildcards: false,
+			},
+		},
+		{
+			name:       "With wildcards",
+			nicsEnvVal: "eth*,wlan0",
+			expectedCfg: &NicConfig{
+				nics:         map[string]struct{}{"eth*": {}, "wlan0": {}},
+				isBlacklist:  false,
+				hasWildcards: true,
+			},
+		},
+		{
+			name:       "Blacklist with wildcards",
+			nicsEnvVal: "-eth*,wlan0",
+			expectedCfg: &NicConfig{
+				nics:         map[string]struct{}{"eth*": {}, "wlan0": {}},
+				isBlacklist:  true,
+				hasWildcards: true,
+			},
+		},
+		{
+			name:       "With whitespace",
+			nicsEnvVal: "eth0, wlan0 , eth1",
+			expectedCfg: &NicConfig{
+				nics:         map[string]struct{}{"eth0": {}, "wlan0": {}, "eth1": {}},
+				isBlacklist:  false,
+				hasWildcards: false,
+			},
+		},
+		{
+			name:       "Only wildcards",
+			nicsEnvVal: "eth*,wlan*",
+			expectedCfg: &NicConfig{
+				nics:         map[string]struct{}{"eth*": {}, "wlan*": {}},
+				isBlacklist:  false,
+				hasWildcards: true,
+			},
+		},
+		{
+			name:       "Leading dash only",
+			nicsEnvVal: "-",
+			expectedCfg: &NicConfig{
+				nics:         map[string]struct{}{},
+				isBlacklist:  true,
+				hasWildcards: false,
+			},
+		},
+		{
+			name:       "Mixed exact and wildcard",
+			nicsEnvVal: "eth0,br-*",
+			expectedCfg: &NicConfig{
+				nics:         map[string]struct{}{"eth0": {}, "br-*": {}},
+				isBlacklist:  false,
+				hasWildcards: true,
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			cfg := newNicConfig(tt.nicsEnvVal)
+			require.NotNil(t, cfg)
+			assert.Equal(t, tt.expectedCfg.isBlacklist, cfg.isBlacklist)
+			assert.Equal(t, tt.expectedCfg.hasWildcards, cfg.hasWildcards)
+			assert.Equal(t, tt.expectedCfg.nics, cfg.nics)
+		})
+	}
+}
+func TestEnsureNetworkInterfacesMap(t *testing.T) {
+	var a Agent
+	var stats system.Stats
+
+	// Initially nil
+	assert.Nil(t, stats.NetworkInterfaces)
+	// Ensure map is created
+	a.ensureNetworkInterfacesMap(&stats)
+	assert.NotNil(t, stats.NetworkInterfaces)
+	// Idempotent
+	a.ensureNetworkInterfacesMap(&stats)
+	assert.NotNil(t, stats.NetworkInterfaces)
+}
+
+func TestLoadAndTickNetBaseline(t *testing.T) {
+	a := &Agent{netIoStats: make(map[uint16]system.NetIoStats)}
+
+	// First call initializes time and returns 0 elapsed
+	ni, elapsed := a.loadAndTickNetBaseline(100)
+	assert.Equal(t, uint64(0), elapsed)
+	assert.False(t, ni.Time.IsZero())
+
+	// Store back what loadAndTick returns to mimic updateNetworkStats behavior
+	a.netIoStats[100] = ni
+
+	time.Sleep(2 * time.Millisecond)
+
+	// Next call should produce >= 0 elapsed and update time
+	ni2, elapsed2 := a.loadAndTickNetBaseline(100)
+	assert.True(t, elapsed2 > 0)
+	assert.False(t, ni2.Time.IsZero())
+}
+
+func TestComputeBytesPerSecond(t *testing.T) {
+	a := &Agent{}
+
+	// No elapsed -> zero rate
+	bytesUp, bytesDown := a.computeBytesPerSecond(0, 2000, 3000, system.NetIoStats{BytesSent: 1000, BytesRecv: 1000})
+	assert.Equal(t, uint64(0), bytesUp)
+	assert.Equal(t, uint64(0), bytesDown)
+
+	// With elapsed -> per-second calculation
+	bytesUp, bytesDown = a.computeBytesPerSecond(500, 6000, 11000, system.NetIoStats{BytesSent: 1000, BytesRecv: 1000})
+	// (6000-1000)*1000/500 = 10000; (11000-1000)*1000/500 = 20000
+	assert.Equal(t, uint64(10000), bytesUp)
+	assert.Equal(t, uint64(20000), bytesDown)
+}
+
+func TestSumAndTrackPerNicDeltas(t *testing.T) {
+	a := &Agent{
+		netInterfaces:             map[string]struct{}{"eth0": {}, "wlan0": {}},
+		netInterfaceDeltaTrackers: make(map[uint16]*deltatracker.DeltaTracker[string, uint64]),
+	}
+
+	// Two samples for same cache interval to verify delta behavior
+	cache := uint16(42)
+	net1 := []psutilNet.IOCountersStat{{Name: "eth0", BytesSent: 1000, BytesRecv: 2000}}
+	stats1 := &system.Stats{}
+	a.ensureNetworkInterfacesMap(stats1)
+	tx1, rx1 := a.sumAndTrackPerNicDeltas(cache, 0, net1, stats1)
+	assert.Equal(t, uint64(1000), tx1)
+	assert.Equal(t, uint64(2000), rx1)
+
+	// Second cycle with elapsed, larger counters -> deltas computed inside
+	net2 := []psutilNet.IOCountersStat{{Name: "eth0", BytesSent: 4000, BytesRecv: 9000}}
+	stats := &system.Stats{}
+	a.ensureNetworkInterfacesMap(stats)
+	tx2, rx2 := a.sumAndTrackPerNicDeltas(cache, 1000, net2, stats)
+	assert.Equal(t, uint64(4000), tx2)
+	assert.Equal(t, uint64(9000), rx2)
+	// Up/Down deltas per second should be (4000-1000)/1s = 3000 and (9000-2000)/1s = 7000
+	ni, ok := stats.NetworkInterfaces["eth0"]
+	assert.True(t, ok)
+	assert.Equal(t, uint64(3000), ni[0])
+	assert.Equal(t, uint64(7000), ni[1])
+}
+
+func TestSumAndTrackPerNicDeltasHandlesCounterReset(t *testing.T) {
+	a := &Agent{
+		netInterfaces:             map[string]struct{}{"eth0": {}},
+		netInterfaceDeltaTrackers: make(map[uint16]*deltatracker.DeltaTracker[string, uint64]),
+	}
+
+	cache := uint16(77)
+
+	// First interval establishes baseline values
+	initial := []psutilNet.IOCountersStat{{Name: "eth0", BytesSent: 4_000, BytesRecv: 6_000}}
+	statsInitial := &system.Stats{}
+	a.ensureNetworkInterfacesMap(statsInitial)
+	_, _ = a.sumAndTrackPerNicDeltas(cache, 0, initial, statsInitial)
+
+	// Second interval increments counters normally so previous snapshot gets populated
+	increment := []psutilNet.IOCountersStat{{Name: "eth0", BytesSent: 9_000, BytesRecv: 11_000}}
+	statsIncrement := &system.Stats{}
+	a.ensureNetworkInterfacesMap(statsIncrement)
+	_, _ = a.sumAndTrackPerNicDeltas(cache, 1_000, increment, statsIncrement)
+
+	niIncrement, ok := statsIncrement.NetworkInterfaces["eth0"]
+	require.True(t, ok)
+	assert.Equal(t, uint64(5_000), niIncrement[0])
+	assert.Equal(t, uint64(5_000), niIncrement[1])
+
+	// Third interval simulates counter reset (values drop below previous totals)
+	reset := []psutilNet.IOCountersStat{{Name: "eth0", BytesSent: 1_200, BytesRecv: 1_500}}
+	statsReset := &system.Stats{}
+	a.ensureNetworkInterfacesMap(statsReset)
+	_, _ = a.sumAndTrackPerNicDeltas(cache, 1_000, reset, statsReset)
+
+	niReset, ok := statsReset.NetworkInterfaces["eth0"]
+	require.True(t, ok)
+	assert.Equal(t, uint64(1_200), niReset[0], "upload delta should match new counter value after reset")
+	assert.Equal(t, uint64(1_500), niReset[1], "download delta should match new counter value after reset")
+}
+
+func TestApplyNetworkTotals(t *testing.T) {
+	tests := []struct {
+		name                  string
+		bytesSentPerSecond    uint64
+		bytesRecvPerSecond    uint64
+		totalBytesSent        uint64
+		totalBytesRecv        uint64
+		expectReset           bool
+		expectedNetworkSent   float64
+		expectedNetworkRecv   float64
+		expectedBandwidthSent uint64
+		expectedBandwidthRecv uint64
+	}{
+		{
+			name:                  "Valid network stats - normal values",
+			bytesSentPerSecond:    1000000, // 1 MB/s
+			bytesRecvPerSecond:    2000000, // 2 MB/s
+			totalBytesSent:        10000000,
+			totalBytesRecv:        20000000,
+			expectReset:           false,
+			expectedNetworkSent:   0.95, // ~1 MB/s rounded to 2 decimals
+			expectedNetworkRecv:   1.91, // ~2 MB/s rounded to 2 decimals
+			expectedBandwidthSent: 1000000,
+			expectedBandwidthRecv: 2000000,
+		},
+		{
+			name:               "Invalid network stats - sent exceeds threshold",
+			bytesSentPerSecond: 11000000000, // ~10.5 GB/s > 10 GB/s threshold
+			bytesRecvPerSecond: 1000000,     // 1 MB/s
+			totalBytesSent:     10000000,
+			totalBytesRecv:     20000000,
+			expectReset:        true,
+		},
+		{
+			name:               "Invalid network stats - recv exceeds threshold",
+			bytesSentPerSecond: 1000000,     // 1 MB/s
+			bytesRecvPerSecond: 11000000000, // ~10.5 GB/s > 10 GB/s threshold
+			totalBytesSent:     10000000,
+			totalBytesRecv:     20000000,
+			expectReset:        true,
+		},
+		{
+			name:               "Invalid network stats - both exceed threshold",
+			bytesSentPerSecond: 12000000000, // ~11.4 GB/s
+			bytesRecvPerSecond: 13000000000, // ~12.4 GB/s
+			totalBytesSent:     10000000,
+			totalBytesRecv:     20000000,
+			expectReset:        true,
+		},
+		{
+			name:                  "Valid network stats - at threshold boundary",
+			bytesSentPerSecond:    10485750000, // ~9999.99 MB/s (rounds to 9999.99)
+			bytesRecvPerSecond:    10485750000, // ~9999.99 MB/s (rounds to 9999.99)
+			totalBytesSent:        10000000,
+			totalBytesRecv:        20000000,
+			expectReset:           false,
+			expectedNetworkSent:   9999.99,
+			expectedNetworkRecv:   9999.99,
+			expectedBandwidthSent: 10485750000,
+			expectedBandwidthRecv: 10485750000,
+		},
+		{
+			name:                  "Zero values",
+			bytesSentPerSecond:    0,
+			bytesRecvPerSecond:    0,
+			totalBytesSent:        0,
+			totalBytesRecv:        0,
+			expectReset:           false,
+			expectedNetworkSent:   0.0,
+			expectedNetworkRecv:   0.0,
+			expectedBandwidthSent: 0,
+			expectedBandwidthRecv: 0,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			// Setup agent with initialized maps
+			a := &Agent{
+				netInterfaces:             make(map[string]struct{}),
+				netIoStats:                make(map[uint16]system.NetIoStats),
+				netInterfaceDeltaTrackers: make(map[uint16]*deltatracker.DeltaTracker[string, uint64]),
+			}
+
+			cacheTimeMs := uint16(100)
+			netIO := []psutilNet.IOCountersStat{
+				{Name: "eth0", BytesSent: 1000, BytesRecv: 2000},
+			}
+			systemStats := &system.Stats{}
+			nis := system.NetIoStats{}
+
+			a.applyNetworkTotals(
+				cacheTimeMs,
+				netIO,
+				systemStats,
+				nis,
+				tt.totalBytesSent,
+				tt.totalBytesRecv,
+				tt.bytesSentPerSecond,
+				tt.bytesRecvPerSecond,
+			)
+
+			if tt.expectReset {
+				// Should have reset network tracking state - maps cleared and stats zeroed
+				assert.NotContains(t, a.netIoStats, cacheTimeMs, "cache entry should be cleared after reset")
+				assert.NotContains(t, a.netInterfaceDeltaTrackers, cacheTimeMs, "tracker should be cleared on reset")
+				assert.Zero(t, systemStats.NetworkSent)
+				assert.Zero(t, systemStats.NetworkRecv)
+				assert.Zero(t, systemStats.Bandwidth[0])
+				assert.Zero(t, systemStats.Bandwidth[1])
+			} else {
+				// Should have applied stats
+				assert.Equal(t, tt.expectedNetworkSent, systemStats.NetworkSent)
+				assert.Equal(t, tt.expectedNetworkRecv, systemStats.NetworkRecv)
+				assert.Equal(t, tt.expectedBandwidthSent, systemStats.Bandwidth[0])
+				assert.Equal(t, tt.expectedBandwidthRecv, systemStats.Bandwidth[1])
+
+				// Should have updated NetIoStats
+				updatedNis := a.netIoStats[cacheTimeMs]
+				assert.Equal(t, tt.totalBytesSent, updatedNis.BytesSent)
+				assert.Equal(t, tt.totalBytesRecv, updatedNis.BytesRecv)
+			}
+		})
+	}
+}

agent/response.go

@@ -0,0 +1,31 @@
+package agent
+
+import (
+	"github.com/fxamacker/cbor/v2"
+	"github.com/henrygd/beszel/internal/common"
+	"github.com/henrygd/beszel/internal/entities/smart"
+	"github.com/henrygd/beszel/internal/entities/system"
+	"github.com/henrygd/beszel/internal/entities/systemd"
+)
+
+// newAgentResponse creates an AgentResponse using legacy typed fields.
+// This maintains backward compatibility with <= 0.17 hubs that expect specific fields.
+func newAgentResponse(data any, requestID *uint32) common.AgentResponse {
+	response := common.AgentResponse{Id: requestID}
+	switch v := data.(type) {
+	case *system.CombinedData:
+		response.SystemData = v
+	case *common.FingerprintResponse:
+		response.Fingerprint = v
+	case string:
+		response.String = &v
+	case map[string]smart.SmartData:
+		response.SmartData = v
+	case systemd.ServiceDetails:
+		response.ServiceInfo = v
+	default:
+		// For unknown types, use the generic Data field
+		response.Data, _ = cbor.Marshal(data)
+	}
+	return response
+}

agent/sensors.go

@@ -1,7 +1,6 @@
 package agent
 
 import (
-	"beszel/internal/entities/system"
 	"context"
 	"fmt"
 	"log/slog"
@@ -11,6 +10,8 @@ import (
 	"strings"
 	"unicode/utf8"
 
+	"github.com/henrygd/beszel/internal/entities/system"
+
 	"github.com/shirou/gopsutil/v4/common"
 	"github.com/shirou/gopsutil/v4/sensors"
 )

agent/sensors_test.go

@@ -4,12 +4,13 @@
 package agent
 
 import (
-	"beszel/internal/entities/system"
 	"context"
 	"fmt"
 	"os"
 	"testing"
 
+	"github.com/henrygd/beszel/internal/entities/system"
+
 	"github.com/shirou/gopsutil/v4/common"
 	"github.com/shirou/gopsutil/v4/sensors"
 	"github.com/stretchr/testify/assert"

agent/sensors_windows.go

@@ -46,9 +46,10 @@ var lhmFs embed.FS
 var (
 	beszelLhm     *lhmProcess
 	beszelLhmOnce sync.Once
+	useLHM        = os.Getenv("LHM") == "true"
 )
 
-var errNoSensors = errors.New("no sensors found (try running as admin)")
+var errNoSensors = errors.New("no sensors found (try running as admin with LHM=true)")
 
 // newlhmProcess copies the embedded LHM executable to a temporary directory and starts it.
 func newlhmProcess() (*lhmProcess, error) {
@@ -139,7 +140,7 @@ func (lhm *lhmProcess) cleanupProcess() {
 }
 
 func (lhm *lhmProcess) getTemps(ctx context.Context) (temps []sensors.TemperatureStat, err error) {
-	if lhm.stoppedNoSensors {
+	if !useLHM || lhm.stoppedNoSensors {
 		// Fall back to gopsutil if we can't get sensors from LHM
 		return sensors.TemperaturesWithContext(ctx)
 	}
@@ -222,6 +223,10 @@ func getSensorTemps(ctx context.Context) (temps []sensors.TemperatureStat, err e
 		}
 	}()
 
+	if !useLHM {
+		return sensors.TemperaturesWithContext(ctx)
+	}
+
 	// Initialize process once
 	beszelLhmOnce.Do(func() {
 		beszelLhm, err = newlhmProcess()

agent/server.go

@@ -1,9 +1,6 @@
 package agent
 
 import (
-	"beszel"
-	"beszel/internal/common"
-	"beszel/internal/entities/system"
 	"encoding/json"
 	"errors"
 	"fmt"
@@ -14,6 +11,10 @@ import (
 	"strings"
 	"time"
 
+	"github.com/henrygd/beszel"
+	"github.com/henrygd/beszel/internal/common"
+	"github.com/henrygd/beszel/internal/entities/system"
+
 	"github.com/blang/semver"
 	"github.com/fxamacker/cbor/v2"
 	"github.com/gliderlabs/ssh"
@@ -126,15 +127,70 @@ func (a *Agent) handleSession(s ssh.Session) {
 
 	hubVersion := a.getHubVersion(sessionID, sessionCtx)
 
-	stats := a.gatherStats(sessionID)
-
-	err := a.writeToSession(s, stats, hubVersion)
-	if err != nil {
-		slog.Error("Error encoding stats", "err", err, "stats", stats)
-		s.Exit(1)
-	} else {
-		s.Exit(0)
+	// Legacy one-shot behavior for older hubs
+	if hubVersion.LT(beszel.MinVersionAgentResponse) {
+		if err := a.handleLegacyStats(s, hubVersion); err != nil {
+			slog.Error("Error encoding stats", "err", err)
+			s.Exit(1)
+			return
+		}
 	}
+
+	var req common.HubRequest[cbor.RawMessage]
+	if err := cbor.NewDecoder(s).Decode(&req); err != nil {
+		// Fallback to legacy one-shot if the first decode fails
+		if err2 := a.handleLegacyStats(s, hubVersion); err2 != nil {
+			slog.Error("Error encoding stats (fallback)", "err", err2)
+			s.Exit(1)
+			return
+		}
+		s.Exit(0)
+		return
+	}
+	if err := a.handleSSHRequest(s, &req); err != nil {
+		slog.Error("SSH request handling failed", "err", err)
+		s.Exit(1)
+		return
+	}
+	s.Exit(0)
+}
+
+// handleSSHRequest builds a handler context and dispatches to the shared registry
+func (a *Agent) handleSSHRequest(w io.Writer, req *common.HubRequest[cbor.RawMessage]) error {
+	// SSH does not support fingerprint auth action
+	if req.Action == common.CheckFingerprint {
+		return cbor.NewEncoder(w).Encode(common.AgentResponse{Error: "unsupported action"})
+	}
+
+	// responder that writes AgentResponse to stdout
+	// Uses legacy typed fields for backward compatibility with <= 0.17
+	sshResponder := func(data any, requestID *uint32) error {
+		response := newAgentResponse(data, requestID)
+		return cbor.NewEncoder(w).Encode(response)
+	}
+
+	ctx := &HandlerContext{
+		Client:       nil,
+		Agent:        a,
+		Request:      req,
+		RequestID:    nil,
+		HubVerified:  true,
+		SendResponse: sshResponder,
+	}
+
+	if handler, ok := a.handlerRegistry.GetHandler(req.Action); ok {
+		if err := handler.Handle(ctx); err != nil {
+			return cbor.NewEncoder(w).Encode(common.AgentResponse{Error: err.Error()})
+		}
+		return nil
+	}
+	return cbor.NewEncoder(w).Encode(common.AgentResponse{Error: fmt.Sprintf("unknown action: %d", req.Action)})
+}
+
+// handleLegacyStats serves the legacy one-shot stats payload for older hubs
+func (a *Agent) handleLegacyStats(w io.Writer, hubVersion semver.Version) error {
+	stats := a.gatherStats(common.DataRequestOptions{CacheTimeMs: 60_000})
+	return a.writeToSession(w, stats, hubVersion)
 }
 
 // writeToSession encodes and writes system statistics to the session.

agent/server_test.go

@@ -1,8 +1,6 @@
 package agent
 
 import (
-	"beszel/internal/entities/container"
-	"beszel/internal/entities/system"
 	"context"
 	"crypto/ed25519"
 	"encoding/json"
@@ -15,6 +13,9 @@ import (
 	"testing"
 	"time"
 
+	"github.com/henrygd/beszel/internal/entities/container"
+	"github.com/henrygd/beszel/internal/entities/system"
+
 	"github.com/blang/semver"
 	"github.com/fxamacker/cbor/v2"
 	"github.com/gliderlabs/ssh"
@@ -512,7 +513,7 @@ func TestWriteToSessionEncoding(t *testing.T) {
 				err = json.Unmarshal([]byte(encodedData), &decodedJson)
 				assert.Error(t, err, "Should not be valid JSON data")
 
-				assert.Equal(t, testData.Info.Hostname, decodedCbor.Info.Hostname)
+				assert.Equal(t, testData.Details.Hostname, decodedCbor.Details.Hostname)
 				assert.Equal(t, testData.Stats.Cpu, decodedCbor.Stats.Cpu)
 			} else {
 				// Should be JSON - try to decode as JSON
@@ -525,7 +526,7 @@ func TestWriteToSessionEncoding(t *testing.T) {
 				assert.Error(t, err, "Should not be valid CBOR data")
 
 				// Verify the decoded JSON data matches our test data
-				assert.Equal(t, testData.Info.Hostname, decodedJson.Info.Hostname)
+				assert.Equal(t, testData.Details.Hostname, decodedJson.Details.Hostname)
 				assert.Equal(t, testData.Stats.Cpu, decodedJson.Stats.Cpu)
 
 				// Verify it looks like JSON (starts with '{' and contains readable field names)
@@ -549,13 +550,12 @@ func createTestCombinedData() *system.CombinedData {
 			DiskUsed:  549755813888,  // 512GB
 			DiskPct:   50.0,
 		},
+		Details: &system.Details{
+			Hostname: "test-host",
+		},
 		Info: system.Info{
-			Hostname:     "test-host",
-			Cores:        8,
-			CpuModel:     "Test CPU Model",
 			Uptime:       3600,
 			AgentVersion: "0.12.0",
-			Os:           system.Linux,
 		},
 		Containers: []*container.Stats{
 			{

agent/smart_nonwindows.go

@@ -0,0 +1,9 @@
+//go:build !windows
+
+package agent
+
+import "errors"
+
+func ensureEmbeddedSmartctl() (string, error) {
+	return "", errors.ErrUnsupported
+}

agent/smart_test.go

@@ -0,0 +1,815 @@
+//go:build testing
+// +build testing
+
+package agent
+
+import (
+	"errors"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/henrygd/beszel/internal/entities/smart"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestParseSmartForScsi(t *testing.T) {
+	fixturePath := filepath.Join("test-data", "smart", "scsi.json")
+	data, err := os.ReadFile(fixturePath)
+	if err != nil {
+		t.Fatalf("failed reading fixture: %v", err)
+	}
+
+	sm := &SmartManager{
+		SmartDataMap: make(map[string]*smart.SmartData),
+	}
+
+	hasData, exitStatus := sm.parseSmartForScsi(data)
+	if !hasData {
+		t.Fatalf("expected SCSI data to parse successfully")
+	}
+	if exitStatus != 0 {
+		t.Fatalf("expected exit status 0, got %d", exitStatus)
+	}
+
+	deviceData, ok := sm.SmartDataMap["9YHSDH9B"]
+	if !ok {
+		t.Fatalf("expected smart data entry for serial 9YHSDH9B")
+	}
+
+	assert.Equal(t, deviceData.ModelName, "YADRO WUH721414AL4204")
+	assert.Equal(t, deviceData.SerialNumber, "9YHSDH9B")
+	assert.Equal(t, deviceData.FirmwareVersion, "C240")
+	assert.Equal(t, deviceData.DiskName, "/dev/sde")
+	assert.Equal(t, deviceData.DiskType, "scsi")
+	assert.EqualValues(t, deviceData.Temperature, 34)
+	assert.Equal(t, deviceData.SmartStatus, "PASSED")
+	assert.EqualValues(t, deviceData.Capacity, 14000519643136)
+
+	if len(deviceData.Attributes) == 0 {
+		t.Fatalf("expected attributes to be populated")
+	}
+
+	assertAttrValue(t, deviceData.Attributes, "PowerOnHours", 458)
+	assertAttrValue(t, deviceData.Attributes, "PowerOnMinutes", 25)
+	assertAttrValue(t, deviceData.Attributes, "GrownDefectList", 0)
+	assertAttrValue(t, deviceData.Attributes, "StartStopCycles", 2)
+	assertAttrValue(t, deviceData.Attributes, "LoadUnloadCycles", 418)
+	assertAttrValue(t, deviceData.Attributes, "ReadGigabytesProcessed", 3641)
+	assertAttrValue(t, deviceData.Attributes, "WriteGigabytesProcessed", 2124590)
+	assertAttrValue(t, deviceData.Attributes, "VerifyGigabytesProcessed", 0)
+}
+
+func TestParseSmartForSata(t *testing.T) {
+	fixturePath := filepath.Join("test-data", "smart", "sda.json")
+	data, err := os.ReadFile(fixturePath)
+	require.NoError(t, err)
+
+	sm := &SmartManager{
+		SmartDataMap: make(map[string]*smart.SmartData),
+	}
+
+	hasData, exitStatus := sm.parseSmartForSata(data)
+	require.True(t, hasData)
+	assert.Equal(t, 64, exitStatus)
+
+	deviceData, ok := sm.SmartDataMap["9C40918040082"]
+	require.True(t, ok, "expected smart data entry for serial 9C40918040082")
+
+	assert.Equal(t, "P3-2TB", deviceData.ModelName)
+	assert.Equal(t, "X0104A0", deviceData.FirmwareVersion)
+	assert.Equal(t, "/dev/sda", deviceData.DiskName)
+	assert.Equal(t, "sat", deviceData.DiskType)
+	assert.Equal(t, uint8(31), deviceData.Temperature)
+	assert.Equal(t, "PASSED", deviceData.SmartStatus)
+	assert.Equal(t, uint64(2048408248320), deviceData.Capacity)
+	if assert.NotEmpty(t, deviceData.Attributes) {
+		assertAttrValue(t, deviceData.Attributes, "Temperature_Celsius", 31)
+	}
+}
+
+func TestParseSmartForSataParentheticalRawValue(t *testing.T) {
+	jsonPayload := []byte(`{
+		"smartctl": {"exit_status": 0},
+		"device": {"name": "/dev/sdz", "type": "sat"},
+		"model_name": "Example",
+		"serial_number": "PARENTHESES123",
+		"firmware_version": "1.0",
+		"user_capacity": {"bytes": 1024},
+		"smart_status": {"passed": true},
+		"temperature": {"current": 25},
+		"ata_smart_attributes": {
+			"table": [
+				{
+					"id": 9,
+					"name": "Power_On_Hours",
+					"value": 93,
+					"worst": 55,
+					"thresh": 0,
+					"when_failed": "",
+                    "raw": {
+                        "value": 57891864217128,
+                        "string": "39925 (212 206 0)"
+                    }
+				}
+			]
+		}
+	}`)
+
+	sm := &SmartManager{SmartDataMap: make(map[string]*smart.SmartData)}
+
+	hasData, exitStatus := sm.parseSmartForSata(jsonPayload)
+	require.True(t, hasData)
+	assert.Equal(t, 0, exitStatus)
+
+	data, ok := sm.SmartDataMap["PARENTHESES123"]
+	require.True(t, ok)
+	require.Len(t, data.Attributes, 1)
+
+	attr := data.Attributes[0]
+	assert.Equal(t, uint64(39925), attr.RawValue)
+	assert.Equal(t, "39925 (212 206 0)", attr.RawString)
+}
+
+func TestParseSmartForNvme(t *testing.T) {
+	fixturePath := filepath.Join("test-data", "smart", "nvme0.json")
+	data, err := os.ReadFile(fixturePath)
+	require.NoError(t, err)
+
+	sm := &SmartManager{
+		SmartDataMap: make(map[string]*smart.SmartData),
+	}
+
+	hasData, exitStatus := sm.parseSmartForNvme(data)
+	require.True(t, hasData)
+	assert.Equal(t, 0, exitStatus)
+
+	deviceData, ok := sm.SmartDataMap["2024031600129"]
+	require.True(t, ok, "expected smart data entry for serial 2024031600129")
+
+	assert.Equal(t, "PELADN 512GB", deviceData.ModelName)
+	assert.Equal(t, "VC2S038E", deviceData.FirmwareVersion)
+	assert.Equal(t, "/dev/nvme0", deviceData.DiskName)
+	assert.Equal(t, "nvme", deviceData.DiskType)
+	assert.Equal(t, uint8(61), deviceData.Temperature)
+	assert.Equal(t, "PASSED", deviceData.SmartStatus)
+	assert.Equal(t, uint64(512110190592), deviceData.Capacity)
+	if assert.NotEmpty(t, deviceData.Attributes) {
+		assertAttrValue(t, deviceData.Attributes, "PercentageUsed", 0)
+		assertAttrValue(t, deviceData.Attributes, "DataUnitsWritten", 16040567)
+	}
+}
+
+func TestHasDataForDevice(t *testing.T) {
+	sm := &SmartManager{
+		SmartDataMap: map[string]*smart.SmartData{
+			"serial-1": {DiskName: "/dev/sda"},
+			"serial-2": nil,
+		},
+	}
+
+	assert.True(t, sm.hasDataForDevice("/dev/sda"))
+	assert.False(t, sm.hasDataForDevice("/dev/sdb"))
+}
+
+func TestDevicesSnapshotReturnsCopy(t *testing.T) {
+	originalDevice := &DeviceInfo{Name: "/dev/sda"}
+	sm := &SmartManager{
+		SmartDevices: []*DeviceInfo{
+			originalDevice,
+			{Name: "/dev/sdb"},
+		},
+	}
+
+	snapshot := sm.devicesSnapshot()
+	require.Len(t, snapshot, 2)
+
+	sm.SmartDevices[0] = &DeviceInfo{Name: "/dev/sdz"}
+	assert.Equal(t, "/dev/sda", snapshot[0].Name)
+
+	snapshot[1] = &DeviceInfo{Name: "/dev/nvme0"}
+	assert.Equal(t, "/dev/sdb", sm.SmartDevices[1].Name)
+
+	sm.SmartDevices = append(sm.SmartDevices, &DeviceInfo{Name: "/dev/nvme1"})
+	assert.Len(t, snapshot, 2)
+}
+
+func TestScanDevicesWithEnvOverride(t *testing.T) {
+	t.Setenv("SMART_DEVICES", "/dev/sda:sat, /dev/nvme0:nvme")
+
+	sm := &SmartManager{
+		SmartDataMap: make(map[string]*smart.SmartData),
+	}
+
+	err := sm.ScanDevices(true)
+	require.NoError(t, err)
+
+	require.Len(t, sm.SmartDevices, 2)
+	assert.Equal(t, "/dev/sda", sm.SmartDevices[0].Name)
+	assert.Equal(t, "sat", sm.SmartDevices[0].Type)
+	assert.Equal(t, "/dev/nvme0", sm.SmartDevices[1].Name)
+	assert.Equal(t, "nvme", sm.SmartDevices[1].Type)
+}
+
+func TestScanDevicesWithEnvOverrideInvalid(t *testing.T) {
+	t.Setenv("SMART_DEVICES", ":sat")
+
+	sm := &SmartManager{
+		SmartDataMap: make(map[string]*smart.SmartData),
+	}
+
+	err := sm.ScanDevices(true)
+	require.Error(t, err)
+}
+
+func TestScanDevicesWithEnvOverrideEmpty(t *testing.T) {
+	t.Setenv("SMART_DEVICES", "   ")
+
+	sm := &SmartManager{
+		SmartDataMap: make(map[string]*smart.SmartData),
+	}
+
+	err := sm.ScanDevices(true)
+	assert.ErrorIs(t, err, errNoValidSmartData)
+	assert.Empty(t, sm.SmartDevices)
+}
+
+func TestSmartctlArgsWithoutType(t *testing.T) {
+	device := &DeviceInfo{Name: "/dev/sda"}
+
+	sm := &SmartManager{}
+
+	args := sm.smartctlArgs(device, true)
+	assert.Equal(t, []string{"-a", "--json=c", "-n", "standby", "/dev/sda"}, args)
+}
+
+func TestSmartctlArgs(t *testing.T) {
+	sm := &SmartManager{}
+
+	sataDevice := &DeviceInfo{Name: "/dev/sda", Type: "sat"}
+	assert.Equal(t,
+		[]string{"-d", "sat", "-a", "--json=c", "-n", "standby", "/dev/sda"},
+		sm.smartctlArgs(sataDevice, true),
+	)
+
+	assert.Equal(t,
+		[]string{"-d", "sat", "-a", "--json=c", "/dev/sda"},
+		sm.smartctlArgs(sataDevice, false),
+	)
+
+	assert.Equal(t,
+		[]string{"-a", "--json=c", "-n", "standby"},
+		sm.smartctlArgs(nil, true),
+	)
+}
+
+func TestResolveRefreshError(t *testing.T) {
+	scanErr := errors.New("scan failed")
+	collectErr := errors.New("collect failed")
+
+	tests := []struct {
+		name        string
+		devices     []*DeviceInfo
+		data        map[string]*smart.SmartData
+		scanErr     error
+		collectErr  error
+		expectedErr error
+		expectNoErr bool
+	}{
+		{
+			name:        "no devices returns scan error",
+			devices:     nil,
+			data:        make(map[string]*smart.SmartData),
+			scanErr:     scanErr,
+			expectedErr: scanErr,
+		},
+		{
+			name:        "has data ignores errors",
+			devices:     []*DeviceInfo{{Name: "/dev/sda"}},
+			data:        map[string]*smart.SmartData{"serial": {}},
+			scanErr:     scanErr,
+			collectErr:  collectErr,
+			expectNoErr: true,
+		},
+		{
+			name:        "collect error preferred",
+			devices:     []*DeviceInfo{{Name: "/dev/sda"}},
+			data:        make(map[string]*smart.SmartData),
+			collectErr:  collectErr,
+			expectedErr: collectErr,
+		},
+		{
+			name:        "scan error returned when no data",
+			devices:     []*DeviceInfo{{Name: "/dev/sda"}},
+			data:        make(map[string]*smart.SmartData),
+			scanErr:     scanErr,
+			expectedErr: scanErr,
+		},
+		{
+			name:        "no errors returns sentinel",
+			devices:     []*DeviceInfo{{Name: "/dev/sda"}},
+			data:        make(map[string]*smart.SmartData),
+			expectedErr: errNoValidSmartData,
+		},
+		{
+			name:        "no devices collect error",
+			devices:     nil,
+			data:        make(map[string]*smart.SmartData),
+			collectErr:  collectErr,
+			expectedErr: collectErr,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			sm := &SmartManager{
+				SmartDevices: tt.devices,
+				SmartDataMap: tt.data,
+			}
+
+			err := sm.resolveRefreshError(tt.scanErr, tt.collectErr)
+			if tt.expectNoErr {
+				assert.NoError(t, err)
+				return
+			}
+
+			if tt.expectedErr == nil {
+				assert.NoError(t, err)
+			} else {
+				assert.Equal(t, tt.expectedErr, err)
+			}
+		})
+	}
+}
+
+func TestParseScan(t *testing.T) {
+	sm := &SmartManager{
+		SmartDataMap: map[string]*smart.SmartData{
+			"serial-active": {DiskName: "/dev/sda"},
+			"serial-stale":  {DiskName: "/dev/sdb"},
+		},
+	}
+
+	scanJSON := []byte(`{
+        "devices": [
+            {"name": "/dev/sda", "type": "sat", "info_name": "/dev/sda [SAT]", "protocol": "ATA"},
+            {"name": "/dev/nvme0", "type": "nvme", "info_name": "/dev/nvme0", "protocol": "NVMe"}
+        ]
+    }`)
+
+	devices, hasData := sm.parseScan(scanJSON)
+	assert.True(t, hasData)
+
+	sm.updateSmartDevices(devices)
+
+	require.Len(t, sm.SmartDevices, 2)
+	assert.Equal(t, "/dev/sda", sm.SmartDevices[0].Name)
+	assert.Equal(t, "sat", sm.SmartDevices[0].Type)
+	assert.Equal(t, "/dev/nvme0", sm.SmartDevices[1].Name)
+	assert.Equal(t, "nvme", sm.SmartDevices[1].Type)
+
+	_, activeExists := sm.SmartDataMap["serial-active"]
+	assert.True(t, activeExists, "active smart data should be preserved when device path remains")
+
+	_, staleExists := sm.SmartDataMap["serial-stale"]
+	assert.False(t, staleExists, "stale smart data entry should be removed when device path disappears")
+}
+
+func TestMergeDeviceListsPrefersConfigured(t *testing.T) {
+	scanned := []*DeviceInfo{
+		{Name: "/dev/sda", Type: "sat", InfoName: "scan-info", Protocol: "ATA"},
+		{Name: "/dev/nvme0", Type: "nvme"},
+	}
+
+	configured := []*DeviceInfo{
+		{Name: "/dev/sda", Type: "sat-override"},
+		{Name: "/dev/sdb", Type: "sat"},
+	}
+
+	merged := mergeDeviceLists(nil, scanned, configured)
+	require.Len(t, merged, 3)
+
+	byName := make(map[string]*DeviceInfo, len(merged))
+	for _, dev := range merged {
+		byName[dev.Name] = dev
+	}
+
+	require.Contains(t, byName, "/dev/sda")
+	assert.Equal(t, "sat-override", byName["/dev/sda"].Type, "configured type should override scanned type")
+	assert.Equal(t, "scan-info", byName["/dev/sda"].InfoName, "scan metadata should be preserved when config does not provide it")
+
+	require.Contains(t, byName, "/dev/nvme0")
+	assert.Equal(t, "nvme", byName["/dev/nvme0"].Type)
+
+	require.Contains(t, byName, "/dev/sdb")
+	assert.Equal(t, "sat", byName["/dev/sdb"].Type)
+}
+
+func TestMergeDeviceListsPreservesVerification(t *testing.T) {
+	existing := []*DeviceInfo{
+		{Name: "/dev/sda", Type: "sat+megaraid", parserType: "sat", typeVerified: true},
+	}
+
+	scanned := []*DeviceInfo{
+		{Name: "/dev/sda", Type: "nvme"},
+	}
+
+	merged := mergeDeviceLists(existing, scanned, nil)
+	require.Len(t, merged, 1)
+
+	device := merged[0]
+	assert.True(t, device.typeVerified)
+	assert.Equal(t, "sat", device.parserType)
+	assert.Equal(t, "sat+megaraid", device.Type)
+}
+
+func TestMergeDeviceListsUpdatesTypeWhenUnverified(t *testing.T) {
+	existing := []*DeviceInfo{
+		{Name: "/dev/sda", Type: "sat", parserType: "sat", typeVerified: false},
+	}
+
+	scanned := []*DeviceInfo{
+		{Name: "/dev/sda", Type: "nvme"},
+	}
+
+	merged := mergeDeviceLists(existing, scanned, nil)
+	require.Len(t, merged, 1)
+
+	device := merged[0]
+	assert.False(t, device.typeVerified)
+	assert.Equal(t, "nvme", device.Type)
+	assert.Equal(t, "", device.parserType)
+}
+
+func TestParseSmartOutputMarksVerified(t *testing.T) {
+	fixturePath := filepath.Join("test-data", "smart", "nvme0.json")
+	data, err := os.ReadFile(fixturePath)
+	require.NoError(t, err)
+
+	sm := &SmartManager{SmartDataMap: make(map[string]*smart.SmartData)}
+	device := &DeviceInfo{Name: "/dev/nvme0"}
+
+	require.True(t, sm.parseSmartOutput(device, data))
+	assert.Equal(t, "nvme", device.Type)
+	assert.Equal(t, "nvme", device.parserType)
+	assert.True(t, device.typeVerified)
+}
+
+func TestParseSmartOutputKeepsCustomType(t *testing.T) {
+	fixturePath := filepath.Join("test-data", "smart", "sda.json")
+	data, err := os.ReadFile(fixturePath)
+	require.NoError(t, err)
+
+	sm := &SmartManager{SmartDataMap: make(map[string]*smart.SmartData)}
+	device := &DeviceInfo{Name: "/dev/sda", Type: "sat+megaraid"}
+
+	require.True(t, sm.parseSmartOutput(device, data))
+	assert.Equal(t, "sat+megaraid", device.Type)
+	assert.Equal(t, "sat", device.parserType)
+	assert.True(t, device.typeVerified)
+}
+
+func TestParseSmartOutputResetsVerificationOnFailure(t *testing.T) {
+	sm := &SmartManager{SmartDataMap: make(map[string]*smart.SmartData)}
+	device := &DeviceInfo{Name: "/dev/sda", Type: "sat", parserType: "sat", typeVerified: true}
+
+	assert.False(t, sm.parseSmartOutput(device, []byte("not json")))
+	assert.False(t, device.typeVerified)
+	assert.Equal(t, "sat", device.parserType)
+}
+
+func assertAttrValue(t *testing.T, attributes []*smart.SmartAttribute, name string, expected uint64) {
+	t.Helper()
+	attr := findAttr(attributes, name)
+	if attr == nil {
+		t.Fatalf("expected attribute %s to be present", name)
+	}
+	if attr.RawValue != expected {
+		t.Fatalf("unexpected attribute %s value: got %d, want %d", name, attr.RawValue, expected)
+	}
+}
+
+func findAttr(attributes []*smart.SmartAttribute, name string) *smart.SmartAttribute {
+	for _, attr := range attributes {
+		if attr != nil && attr.Name == name {
+			return attr
+		}
+	}
+	return nil
+}
+
+func TestIsVirtualDevice(t *testing.T) {
+	sm := &SmartManager{}
+
+	tests := []struct {
+		name     string
+		vendor   string
+		product  string
+		model    string
+		expected bool
+	}{
+		{"regular drive", "SEAGATE", "ST1000DM003", "ST1000DM003-1CH162", false},
+		{"qemu virtual", "QEMU", "QEMU HARDDISK", "QEMU HARDDISK", true},
+		{"virtualbox virtual", "VBOX", "HARDDISK", "VBOX HARDDISK", true},
+		{"vmware virtual", "VMWARE", "Virtual disk", "VMWARE Virtual disk", true},
+		{"virtual in model", "ATA", "VIRTUAL", "VIRTUAL DISK", true},
+		{"iet virtual", "IET", "VIRTUAL-DISK", "VIRTUAL-DISK", true},
+		{"hyper-v virtual", "MSFT", "VIRTUAL HD", "VIRTUAL HD", true},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			data := &smart.SmartInfoForSata{
+				ScsiVendor:  tt.vendor,
+				ScsiProduct: tt.product,
+				ModelName:   tt.model,
+			}
+			result := sm.isVirtualDevice(data)
+			assert.Equal(t, tt.expected, result)
+		})
+	}
+}
+
+func TestIsVirtualDeviceNvme(t *testing.T) {
+	sm := &SmartManager{}
+
+	tests := []struct {
+		name     string
+		model    string
+		expected bool
+	}{
+		{"regular nvme", "Samsung SSD 970 EVO Plus 1TB", false},
+		{"qemu virtual", "QEMU NVMe Ctrl", true},
+		{"virtualbox virtual", "VBOX NVMe", true},
+		{"vmware virtual", "VMWARE NVMe", true},
+		{"virtual in model", "Virtual NVMe Device", true},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			data := &smart.SmartInfoForNvme{
+				ModelName: tt.model,
+			}
+			result := sm.isVirtualDeviceNvme(data)
+			assert.Equal(t, tt.expected, result)
+		})
+	}
+}
+
+func TestIsVirtualDeviceScsi(t *testing.T) {
+	sm := &SmartManager{}
+
+	tests := []struct {
+		name     string
+		vendor   string
+		product  string
+		model    string
+		expected bool
+	}{
+		{"regular scsi", "SEAGATE", "ST1000DM003", "ST1000DM003-1CH162", false},
+		{"qemu virtual", "QEMU", "QEMU HARDDISK", "QEMU HARDDISK", true},
+		{"virtualbox virtual", "VBOX", "HARDDISK", "VBOX HARDDISK", true},
+		{"vmware virtual", "VMWARE", "Virtual disk", "VMWARE Virtual disk", true},
+		{"virtual in model", "ATA", "VIRTUAL", "VIRTUAL DISK", true},
+		{"iet virtual", "IET", "VIRTUAL-DISK", "VIRTUAL-DISK", true},
+		{"hyper-v virtual", "MSFT", "VIRTUAL HD", "VIRTUAL HD", true},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			data := &smart.SmartInfoForScsi{
+				ScsiVendor:    tt.vendor,
+				ScsiProduct:   tt.product,
+				ScsiModelName: tt.model,
+			}
+			result := sm.isVirtualDeviceScsi(data)
+			assert.Equal(t, tt.expected, result)
+		})
+	}
+}
+
+func TestRefreshExcludedDevices(t *testing.T) {
+	tests := []struct {
+		name         string
+		envValue     string
+		expectedDevs map[string]struct{}
+	}{
+		{
+			name:         "empty env",
+			envValue:     "",
+			expectedDevs: map[string]struct{}{},
+		},
+		{
+			name:     "single device",
+			envValue: "/dev/sda",
+			expectedDevs: map[string]struct{}{
+				"/dev/sda": {},
+			},
+		},
+		{
+			name:     "multiple devices",
+			envValue: "/dev/sda,/dev/sdb,/dev/nvme0",
+			expectedDevs: map[string]struct{}{
+				"/dev/sda":   {},
+				"/dev/sdb":   {},
+				"/dev/nvme0": {},
+			},
+		},
+		{
+			name:     "devices with whitespace",
+			envValue: " /dev/sda , /dev/sdb ,  /dev/nvme0  ",
+			expectedDevs: map[string]struct{}{
+				"/dev/sda":   {},
+				"/dev/sdb":   {},
+				"/dev/nvme0": {},
+			},
+		},
+		{
+			name:     "duplicate devices",
+			envValue: "/dev/sda,/dev/sdb,/dev/sda",
+			expectedDevs: map[string]struct{}{
+				"/dev/sda": {},
+				"/dev/sdb": {},
+			},
+		},
+		{
+			name:     "empty entries and whitespace",
+			envValue: "/dev/sda,, /dev/sdb , , ",
+			expectedDevs: map[string]struct{}{
+				"/dev/sda": {},
+				"/dev/sdb": {},
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			if tt.envValue != "" {
+				t.Setenv("EXCLUDE_SMART", tt.envValue)
+			} else {
+				// Ensure env var is not set for empty test
+				os.Unsetenv("EXCLUDE_SMART")
+			}
+
+			sm := &SmartManager{}
+			sm.refreshExcludedDevices()
+
+			assert.Equal(t, tt.expectedDevs, sm.excludedDevices)
+		})
+	}
+}
+
+func TestIsExcludedDevice(t *testing.T) {
+	sm := &SmartManager{
+		excludedDevices: map[string]struct{}{
+			"/dev/sda":   {},
+			"/dev/nvme0": {},
+		},
+	}
+
+	tests := []struct {
+		name         string
+		deviceName   string
+		expectedBool bool
+	}{
+		{"excluded device sda", "/dev/sda", true},
+		{"excluded device nvme0", "/dev/nvme0", true},
+		{"non-excluded device sdb", "/dev/sdb", false},
+		{"non-excluded device nvme1", "/dev/nvme1", false},
+		{"empty device name", "", false},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			result := sm.isExcludedDevice(tt.deviceName)
+			assert.Equal(t, tt.expectedBool, result)
+		})
+	}
+}
+
+func TestFilterExcludedDevices(t *testing.T) {
+	tests := []struct {
+		name           string
+		excludedDevs   map[string]struct{}
+		inputDevices   []*DeviceInfo
+		expectedDevs   []*DeviceInfo
+		expectedLength int
+	}{
+		{
+			name:         "no exclusions",
+			excludedDevs: map[string]struct{}{},
+			inputDevices: []*DeviceInfo{
+				{Name: "/dev/sda"},
+				{Name: "/dev/sdb"},
+				{Name: "/dev/nvme0"},
+			},
+			expectedDevs: []*DeviceInfo{
+				{Name: "/dev/sda"},
+				{Name: "/dev/sdb"},
+				{Name: "/dev/nvme0"},
+			},
+			expectedLength: 3,
+		},
+		{
+			name: "some devices excluded",
+			excludedDevs: map[string]struct{}{
+				"/dev/sda":   {},
+				"/dev/nvme0": {},
+			},
+			inputDevices: []*DeviceInfo{
+				{Name: "/dev/sda"},
+				{Name: "/dev/sdb"},
+				{Name: "/dev/nvme0"},
+				{Name: "/dev/nvme1"},
+			},
+			expectedDevs: []*DeviceInfo{
+				{Name: "/dev/sdb"},
+				{Name: "/dev/nvme1"},
+			},
+			expectedLength: 2,
+		},
+		{
+			name: "all devices excluded",
+			excludedDevs: map[string]struct{}{
+				"/dev/sda": {},
+				"/dev/sdb": {},
+			},
+			inputDevices: []*DeviceInfo{
+				{Name: "/dev/sda"},
+				{Name: "/dev/sdb"},
+			},
+			expectedDevs:   []*DeviceInfo{},
+			expectedLength: 0,
+		},
+		{
+			name:           "nil devices",
+			excludedDevs:   map[string]struct{}{},
+			inputDevices:   nil,
+			expectedDevs:   []*DeviceInfo{},
+			expectedLength: 0,
+		},
+		{
+			name: "filter nil and empty name devices",
+			excludedDevs: map[string]struct{}{
+				"/dev/sda": {},
+			},
+			inputDevices: []*DeviceInfo{
+				{Name: "/dev/sda"},
+				nil,
+				{Name: ""},
+				{Name: "/dev/sdb"},
+			},
+			expectedDevs: []*DeviceInfo{
+				{Name: "/dev/sdb"},
+			},
+			expectedLength: 1,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			sm := &SmartManager{
+				excludedDevices: tt.excludedDevs,
+			}
+
+			result := sm.filterExcludedDevices(tt.inputDevices)
+
+			assert.Len(t, result, tt.expectedLength)
+			assert.Equal(t, tt.expectedDevs, result)
+		})
+	}
+}
+
+func TestIsNvmeControllerPath(t *testing.T) {
+	tests := []struct {
+		path     string
+		expected bool
+	}{
+		// Controller paths (should return true)
+		{"/dev/nvme0", true},
+		{"/dev/nvme1", true},
+		{"/dev/nvme10", true},
+		{"nvme0", true},
+
+		// Namespace paths (should return false)
+		{"/dev/nvme0n1", false},
+		{"/dev/nvme1n1", false},
+		{"/dev/nvme0n1p1", false},
+		{"nvme0n1", false},
+
+		// Non-NVMe paths (should return false)
+		{"/dev/sda", false},
+		{"/dev/sda1", false},
+		{"/dev/hda", false},
+		{"", false},
+		{"/dev/nvme", false},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.path, func(t *testing.T) {
+			result := isNvmeControllerPath(tt.path)
+			assert.Equal(t, tt.expected, result, "path: %s", tt.path)
+		})
+	}
+}

agent/smart_windows.go

@@ -0,0 +1,40 @@
+//go:build windows
+
+package agent
+
+import (
+	_ "embed"
+	"fmt"
+	"os"
+	"path/filepath"
+	"sync"
+)
+
+//go:embed smartmontools/smartctl.exe
+var embeddedSmartctl []byte
+
+var (
+	smartctlOnce sync.Once
+	smartctlPath string
+	smartctlErr  error
+)
+
+func ensureEmbeddedSmartctl() (string, error) {
+	smartctlOnce.Do(func() {
+		destDir := filepath.Join(os.TempDir(), "beszel", "smartmontools")
+		if err := os.MkdirAll(destDir, 0o755); err != nil {
+			smartctlErr = fmt.Errorf("failed to create smartctl directory: %w", err)
+			return
+		}
+
+		destPath := filepath.Join(destDir, "smartctl.exe")
+		if err := os.WriteFile(destPath, embeddedSmartctl, 0o755); err != nil {
+			smartctlErr = fmt.Errorf("failed to write embedded smartctl: %w", err)
+			return
+		}
+
+		smartctlPath = destPath
+	})
+
+	return smartctlPath, smartctlErr
+}

agent/system.go

@@ -0,0 +1,298 @@
+package agent
+
+import (
+	"bufio"
+	"errors"
+	"fmt"
+	"log/slog"
+	"os"
+	"runtime"
+	"strconv"
+	"strings"
+	"time"
+
+	"github.com/henrygd/beszel"
+	"github.com/henrygd/beszel/agent/battery"
+	"github.com/henrygd/beszel/internal/entities/container"
+	"github.com/henrygd/beszel/internal/entities/system"
+
+	"github.com/shirou/gopsutil/v4/cpu"
+	"github.com/shirou/gopsutil/v4/host"
+	"github.com/shirou/gopsutil/v4/load"
+	"github.com/shirou/gopsutil/v4/mem"
+)
+
+// prevDisk stores previous per-device disk counters for a given cache interval
+type prevDisk struct {
+	readBytes  uint64
+	writeBytes uint64
+	at         time.Time
+}
+
+// Sets initial / non-changing values about the host system
+func (a *Agent) refreshSystemDetails() {
+	a.systemInfo.AgentVersion = beszel.Version
+
+	// get host info from Docker if available
+	var hostInfo container.HostInfo
+
+	if a.dockerManager != nil {
+		a.systemDetails.Podman = a.dockerManager.IsPodman()
+		hostInfo, _ = a.dockerManager.GetHostInfo()
+	}
+
+	a.systemDetails.Hostname, _ = os.Hostname()
+	if arch, err := host.KernelArch(); err == nil {
+		a.systemDetails.Arch = arch
+	} else {
+		a.systemDetails.Arch = runtime.GOARCH
+	}
+
+	platform, _, version, _ := host.PlatformInformation()
+
+	if platform == "darwin" {
+		a.systemDetails.Os = system.Darwin
+		a.systemDetails.OsName = fmt.Sprintf("macOS %s", version)
+	} else if strings.Contains(platform, "indows") {
+		a.systemDetails.Os = system.Windows
+		a.systemDetails.OsName = strings.Replace(platform, "Microsoft ", "", 1)
+		a.systemDetails.Kernel = version
+	} else if platform == "freebsd" {
+		a.systemDetails.Os = system.Freebsd
+		a.systemDetails.Kernel, _ = host.KernelVersion()
+		if prettyName, err := getOsPrettyName(); err == nil {
+			a.systemDetails.OsName = prettyName
+		} else {
+			a.systemDetails.OsName = "FreeBSD"
+		}
+	} else {
+		a.systemDetails.Os = system.Linux
+		a.systemDetails.OsName = hostInfo.OperatingSystem
+		if a.systemDetails.OsName == "" {
+			if prettyName, err := getOsPrettyName(); err == nil {
+				a.systemDetails.OsName = prettyName
+			} else {
+				a.systemDetails.OsName = platform
+			}
+		}
+		a.systemDetails.Kernel = hostInfo.KernelVersion
+		if a.systemDetails.Kernel == "" {
+			a.systemDetails.Kernel, _ = host.KernelVersion()
+		}
+	}
+
+	// cpu model
+	if info, err := cpu.Info(); err == nil && len(info) > 0 {
+		a.systemDetails.CpuModel = info[0].ModelName
+	}
+	// cores / threads
+	cores, _ := cpu.Counts(false)
+	threads := hostInfo.NCPU
+	if threads == 0 {
+		threads, _ = cpu.Counts(true)
+	}
+	// in lxc, logical cores reflects container limits, so use that as cores if lower
+	if threads > 0 && threads < cores {
+		cores = threads
+	}
+	a.systemDetails.Cores = cores
+	a.systemDetails.Threads = threads
+
+	// total memory
+	a.systemDetails.MemoryTotal = hostInfo.MemTotal
+	if a.systemDetails.MemoryTotal == 0 {
+		if v, err := mem.VirtualMemory(); err == nil {
+			a.systemDetails.MemoryTotal = v.Total
+		}
+	}
+
+	// zfs
+	if _, err := getARCSize(); err != nil {
+		slog.Debug("Not monitoring ZFS ARC", "err", err)
+	} else {
+		a.zfs = true
+	}
+}
+
+// Returns current info, stats about the host system
+func (a *Agent) getSystemStats(cacheTimeMs uint16) system.Stats {
+	var systemStats system.Stats
+
+	// battery
+	if batteryPercent, batteryState, err := battery.GetBatteryStats(); err == nil {
+		systemStats.Battery[0] = batteryPercent
+		systemStats.Battery[1] = batteryState
+	}
+
+	// cpu metrics
+	cpuMetrics, err := getCpuMetrics(cacheTimeMs)
+	if err == nil {
+		systemStats.Cpu = twoDecimals(cpuMetrics.Total)
+		systemStats.CpuBreakdown = []float64{
+			twoDecimals(cpuMetrics.User),
+			twoDecimals(cpuMetrics.System),
+			twoDecimals(cpuMetrics.Iowait),
+			twoDecimals(cpuMetrics.Steal),
+			twoDecimals(cpuMetrics.Idle),
+		}
+	} else {
+		slog.Error("Error getting cpu metrics", "err", err)
+	}
+
+	// per-core cpu usage
+	if perCoreUsage, err := getPerCoreCpuUsage(cacheTimeMs); err == nil {
+		systemStats.CpuCoresUsage = perCoreUsage
+	}
+
+	// load average
+	if avgstat, err := load.Avg(); err == nil {
+		systemStats.LoadAvg[0] = avgstat.Load1
+		systemStats.LoadAvg[1] = avgstat.Load5
+		systemStats.LoadAvg[2] = avgstat.Load15
+		slog.Debug("Load average", "5m", avgstat.Load5, "15m", avgstat.Load15)
+	} else {
+		slog.Error("Error getting load average", "err", err)
+	}
+
+	// memory
+	if v, err := mem.VirtualMemory(); err == nil {
+		// swap
+		systemStats.Swap = bytesToGigabytes(v.SwapTotal)
+		systemStats.SwapUsed = bytesToGigabytes(v.SwapTotal - v.SwapFree - v.SwapCached)
+		// cache + buffers value for default mem calculation
+		// note: gopsutil automatically adds SReclaimable to v.Cached
+		cacheBuff := v.Cached + v.Buffers - v.Shared
+		if cacheBuff <= 0 {
+			cacheBuff = max(v.Total-v.Free-v.Used, 0)
+		}
+		// htop memory calculation overrides (likely outdated as of mid 2025)
+		if a.memCalc == "htop" {
+			// cacheBuff = v.Cached + v.Buffers - v.Shared
+			v.Used = v.Total - (v.Free + cacheBuff)
+			v.UsedPercent = float64(v.Used) / float64(v.Total) * 100.0
+		}
+		// if a.memCalc == "legacy" {
+		// 	v.Used = v.Total - v.Free - v.Buffers - v.Cached
+		// 	cacheBuff = v.Total - v.Free - v.Used
+		// 	v.UsedPercent = float64(v.Used) / float64(v.Total) * 100.0
+		// }
+		// subtract ZFS ARC size from used memory and add as its own category
+		if a.zfs {
+			if arcSize, _ := getARCSize(); arcSize > 0 && arcSize < v.Used {
+				v.Used = v.Used - arcSize
+				v.UsedPercent = float64(v.Used) / float64(v.Total) * 100.0
+				systemStats.MemZfsArc = bytesToGigabytes(arcSize)
+			}
+		}
+		systemStats.Mem = bytesToGigabytes(v.Total)
+		systemStats.MemBuffCache = bytesToGigabytes(cacheBuff)
+		systemStats.MemUsed = bytesToGigabytes(v.Used)
+		systemStats.MemPct = twoDecimals(v.UsedPercent)
+	}
+
+	// disk usage
+	a.updateDiskUsage(&systemStats)
+
+	// disk i/o (cache-aware per interval)
+	a.updateDiskIo(cacheTimeMs, &systemStats)
+
+	// network stats (per cache interval)
+	a.updateNetworkStats(cacheTimeMs, &systemStats)
+
+	// temperatures
+	// TODO: maybe refactor to methods on systemStats
+	a.updateTemperatures(&systemStats)
+
+	// GPU data
+	if a.gpuManager != nil {
+		// reset high gpu percent
+		a.systemInfo.GpuPct = 0
+		// get current GPU data
+		if gpuData := a.gpuManager.GetCurrentData(cacheTimeMs); len(gpuData) > 0 {
+			systemStats.GPUData = gpuData
+
+			// add temperatures
+			if systemStats.Temperatures == nil {
+				systemStats.Temperatures = make(map[string]float64, len(gpuData))
+			}
+			highestTemp := 0.0
+			for _, gpu := range gpuData {
+				if gpu.Temperature > 0 {
+					systemStats.Temperatures[gpu.Name] = gpu.Temperature
+					if a.sensorConfig.primarySensor == gpu.Name {
+						a.systemInfo.DashboardTemp = gpu.Temperature
+					}
+					if gpu.Temperature > highestTemp {
+						highestTemp = gpu.Temperature
+					}
+				}
+				// update high gpu percent for dashboard
+				a.systemInfo.GpuPct = max(a.systemInfo.GpuPct, gpu.Usage)
+			}
+			// use highest temp for dashboard temp if dashboard temp is unset
+			if a.systemInfo.DashboardTemp == 0 {
+				a.systemInfo.DashboardTemp = highestTemp
+			}
+		}
+	}
+
+	// update system info
+	a.systemInfo.ConnectionType = a.connectionManager.ConnectionType
+	a.systemInfo.Cpu = systemStats.Cpu
+	a.systemInfo.LoadAvg = systemStats.LoadAvg
+	a.systemInfo.MemPct = systemStats.MemPct
+	a.systemInfo.DiskPct = systemStats.DiskPct
+	a.systemInfo.Battery = systemStats.Battery
+	a.systemInfo.Uptime, _ = host.Uptime()
+	a.systemInfo.BandwidthBytes = systemStats.Bandwidth[0] + systemStats.Bandwidth[1]
+	a.systemInfo.Threads = a.systemDetails.Threads
+
+	return systemStats
+}
+
+// Returns the size of the ZFS ARC memory cache in bytes
+func getARCSize() (uint64, error) {
+	file, err := os.Open("/proc/spl/kstat/zfs/arcstats")
+	if err != nil {
+		return 0, err
+	}
+	defer file.Close()
+
+	// Scan the lines
+	scanner := bufio.NewScanner(file)
+	for scanner.Scan() {
+		line := scanner.Text()
+		if strings.HasPrefix(line, "size") {
+			// Example line: size 4 15032385536
+			fields := strings.Fields(line)
+			if len(fields) < 3 {
+				return 0, err
+			}
+			// Return the size as uint64
+			return strconv.ParseUint(fields[2], 10, 64)
+		}
+	}
+
+	return 0, fmt.Errorf("failed to parse size field")
+}
+
+// getOsPrettyName attempts to get the pretty OS name from /etc/os-release on Linux systems
+func getOsPrettyName() (string, error) {
+	file, err := os.Open("/etc/os-release")
+	if err != nil {
+		return "", err
+	}
+	defer file.Close()
+
+	scanner := bufio.NewScanner(file)
+	for scanner.Scan() {
+		line := scanner.Text()
+		if after, ok := strings.CutPrefix(line, "PRETTY_NAME="); ok {
+			value := after
+			value = strings.Trim(value, `"`)
+			return value, nil
+		}
+	}
+
+	return "", errors.New("pretty name not found")
+}

agent/systemd.go

@@ -0,0 +1,299 @@
+//go:build linux
+
+package agent
+
+import (
+	"context"
+	"errors"
+	"log/slog"
+	"maps"
+	"math"
+	"os"
+	"strconv"
+	"strings"
+	"sync"
+	"time"
+
+	"github.com/coreos/go-systemd/v22/dbus"
+	"github.com/henrygd/beszel/internal/entities/systemd"
+)
+
+var errNoActiveTime = errors.New("no active time")
+
+// systemdManager manages the collection of systemd service statistics.
+type systemdManager struct {
+	sync.Mutex
+	serviceStatsMap map[string]*systemd.Service
+	isRunning       bool
+	hasFreshStats   bool
+	patterns        []string
+}
+
+// isSystemdAvailable checks if systemd is used on the system to avoid unnecessary connection attempts (#1548)
+func isSystemdAvailable() bool {
+	paths := []string{
+		"/run/systemd/system",
+		"/run/dbus/system_bus_socket",
+		"/var/run/dbus/system_bus_socket",
+	}
+	for _, path := range paths {
+		if _, err := os.Stat(path); err == nil {
+			return true
+		}
+	}
+	if data, err := os.ReadFile("/proc/1/comm"); err == nil {
+		return strings.TrimSpace(string(data)) == "systemd"
+	}
+	return false
+}
+
+// newSystemdManager creates a new systemdManager.
+func newSystemdManager() (*systemdManager, error) {
+	if skipSystemd, _ := GetEnv("SKIP_SYSTEMD"); skipSystemd == "true" {
+		return nil, nil
+	}
+
+	// Check if systemd is available on the system before attempting connection
+	if !isSystemdAvailable() {
+		slog.Debug("Systemd not available")
+		return nil, nil
+	}
+
+	conn, err := dbus.NewSystemConnectionContext(context.Background())
+	if err != nil {
+		slog.Debug("Error connecting to systemd", "err", err, "ref", "https://beszel.dev/guide/systemd")
+		return nil, err
+	}
+
+	manager := &systemdManager{
+		serviceStatsMap: make(map[string]*systemd.Service),
+		patterns:        getServicePatterns(),
+	}
+
+	manager.startWorker(conn)
+
+	return manager, nil
+}
+
+func (sm *systemdManager) startWorker(conn *dbus.Conn) {
+	if sm.isRunning {
+		return
+	}
+	sm.isRunning = true
+	// prime the service stats map with the current services
+	_ = sm.getServiceStats(conn, true)
+	// update the services every 10 minutes
+	go func() {
+		for {
+			time.Sleep(time.Minute * 10)
+			_ = sm.getServiceStats(nil, true)
+		}
+	}()
+}
+
+// getServiceStatsCount returns the number of systemd services.
+func (sm *systemdManager) getServiceStatsCount() int {
+	return len(sm.serviceStatsMap)
+}
+
+// getFailedServiceCount returns the number of systemd services in a failed state.
+func (sm *systemdManager) getFailedServiceCount() uint16 {
+	sm.Lock()
+	defer sm.Unlock()
+	count := uint16(0)
+	for _, service := range sm.serviceStatsMap {
+		if service.State == systemd.StatusFailed {
+			count++
+		}
+	}
+	return count
+}
+
+// getServiceStats collects statistics for all running systemd services.
+func (sm *systemdManager) getServiceStats(conn *dbus.Conn, refresh bool) []*systemd.Service {
+	// start := time.Now()
+	// defer func() {
+	// 	slog.Info("systemdManager.getServiceStats", "duration", time.Since(start))
+	// }()
+
+	var services []*systemd.Service
+	var err error
+
+	if !refresh {
+		// return nil
+		sm.Lock()
+		defer sm.Unlock()
+		for _, service := range sm.serviceStatsMap {
+			services = append(services, service)
+		}
+		sm.hasFreshStats = false
+		return services
+	}
+
+	if conn == nil || !conn.Connected() {
+		conn, err = dbus.NewSystemConnectionContext(context.Background())
+		if err != nil {
+			return nil
+		}
+		defer conn.Close()
+	}
+
+	units, err := conn.ListUnitsByPatternsContext(context.Background(), []string{"loaded"}, sm.patterns)
+	if err != nil {
+		slog.Error("Error listing systemd service units", "err", err)
+		return nil
+	}
+
+	for _, unit := range units {
+		service, err := sm.updateServiceStats(conn, unit)
+		if err != nil {
+			continue
+		}
+		services = append(services, service)
+	}
+	sm.hasFreshStats = true
+	return services
+}
+
+// updateServiceStats updates the statistics for a single systemd service.
+func (sm *systemdManager) updateServiceStats(conn *dbus.Conn, unit dbus.UnitStatus) (*systemd.Service, error) {
+	sm.Lock()
+	defer sm.Unlock()
+
+	ctx := context.Background()
+
+	// if service has never been active (no active since time), skip it
+	if activeEnterTsProp, err := conn.GetUnitTypePropertyContext(ctx, unit.Name, "Unit", "ActiveEnterTimestamp"); err == nil {
+		if ts, ok := activeEnterTsProp.Value.Value().(uint64); !ok || ts == 0 || ts == math.MaxUint64 {
+			return nil, errNoActiveTime
+		}
+	} else {
+		return nil, err
+	}
+
+	service, serviceExists := sm.serviceStatsMap[unit.Name]
+	if !serviceExists {
+		service = &systemd.Service{Name: unescapeServiceName(strings.TrimSuffix(unit.Name, ".service"))}
+		sm.serviceStatsMap[unit.Name] = service
+	}
+
+	memPeak := service.MemPeak
+	if memPeakProp, err := conn.GetUnitTypePropertyContext(ctx, unit.Name, "Service", "MemoryPeak"); err == nil {
+		// If memPeak is MaxUint64 the api is saying it's not available
+		if v, ok := memPeakProp.Value.Value().(uint64); ok && v != math.MaxUint64 {
+			memPeak = v
+		}
+	}
+
+	var memUsage uint64
+	if memProp, err := conn.GetUnitTypePropertyContext(ctx, unit.Name, "Service", "MemoryCurrent"); err == nil {
+		// If memUsage is MaxUint64 the api is saying it's not available
+		if v, ok := memProp.Value.Value().(uint64); ok && v != math.MaxUint64 {
+			memUsage = v
+		}
+	}
+
+	service.State = systemd.ParseServiceStatus(unit.ActiveState)
+	service.Sub = systemd.ParseServiceSubState(unit.SubState)
+
+	// some systems always return 0 for mem peak, so we should update the peak if the current usage is greater
+	if memUsage > memPeak {
+		memPeak = memUsage
+	}
+
+	var cpuUsage uint64
+	if cpuProp, err := conn.GetUnitTypePropertyContext(ctx, unit.Name, "Service", "CPUUsageNSec"); err == nil {
+		if v, ok := cpuProp.Value.Value().(uint64); ok {
+			cpuUsage = v
+		}
+	}
+
+	service.Mem = memUsage
+	if memPeak > service.MemPeak {
+		service.MemPeak = memPeak
+	}
+	service.UpdateCPUPercent(cpuUsage)
+
+	return service, nil
+}
+
+// getServiceDetails collects extended information for a specific systemd service.
+func (sm *systemdManager) getServiceDetails(serviceName string) (systemd.ServiceDetails, error) {
+	conn, err := dbus.NewSystemConnectionContext(context.Background())
+	if err != nil {
+		return nil, err
+	}
+	defer conn.Close()
+
+	unitName := serviceName
+	if !strings.HasSuffix(unitName, ".service") {
+		unitName += ".service"
+	}
+
+	ctx := context.Background()
+	props, err := conn.GetUnitPropertiesContext(ctx, unitName)
+	if err != nil {
+		return nil, err
+	}
+
+	// Start with all unit properties
+	details := make(systemd.ServiceDetails)
+	maps.Copy(details, props)
+
+	// // Add service-specific properties
+	servicePropNames := []string{
+		"MainPID", "ExecMainPID", "TasksCurrent", "TasksMax",
+		"MemoryCurrent", "MemoryPeak", "MemoryLimit", "CPUUsageNSec",
+		"NRestarts", "ExecMainStartTimestampRealtime", "Result",
+	}
+
+	for _, propName := range servicePropNames {
+		if variant, err := conn.GetUnitTypePropertyContext(ctx, unitName, "Service", propName); err == nil {
+			value := variant.Value.Value()
+			// Check if the value is MaxUint64, which indicates unlimited/infinite
+			if uint64Value, ok := value.(uint64); ok && uint64Value == math.MaxUint64 {
+				// Set to nil to indicate unlimited - frontend will handle this appropriately
+				details[propName] = nil
+			} else {
+				details[propName] = value
+			}
+		}
+	}
+
+	return details, nil
+}
+
+// unescapeServiceName unescapes systemd service names that contain C-style escape sequences like \x2d
+func unescapeServiceName(name string) string {
+	if !strings.Contains(name, "\\x") {
+		return name
+	}
+	unescaped, err := strconv.Unquote("\"" + name + "\"")
+	if err != nil {
+		return name
+	}
+	return unescaped
+}
+
+// getServicePatterns returns the list of service patterns to match.
+// It reads from the SERVICE_PATTERNS environment variable if set,
+// otherwise defaults to "*service".
+func getServicePatterns() []string {
+	patterns := []string{}
+	if envPatterns, _ := GetEnv("SERVICE_PATTERNS"); envPatterns != "" {
+		for pattern := range strings.SplitSeq(envPatterns, ",") {
+			pattern = strings.TrimSpace(pattern)
+			if pattern == "" {
+				continue
+			}
+			if !strings.HasSuffix(pattern, ".service") {
+				pattern += ".service"
+			}
+			patterns = append(patterns, pattern)
+		}
+	}
+	if len(patterns) == 0 {
+		patterns = []string{"*.service"}
+	}
+	return patterns
+}

agent/systemd_nonlinux.go

@@ -0,0 +1,38 @@
+//go:build !linux
+
+package agent
+
+import (
+	"errors"
+
+	"github.com/henrygd/beszel/internal/entities/systemd"
+)
+
+// systemdManager manages the collection of systemd service statistics.
+type systemdManager struct {
+	hasFreshStats bool
+}
+
+// newSystemdManager creates a new systemdManager.
+func newSystemdManager() (*systemdManager, error) {
+	return &systemdManager{}, nil
+}
+
+// getServiceStats returns nil for non-linux systems.
+func (sm *systemdManager) getServiceStats(conn any, refresh bool) []*systemd.Service {
+	return nil
+}
+
+// getServiceStatsCount returns 0 for non-linux systems.
+func (sm *systemdManager) getServiceStatsCount() int {
+	return 0
+}
+
+// getFailedServiceCount returns 0 for non-linux systems.
+func (sm *systemdManager) getFailedServiceCount() uint16 {
+	return 0
+}
+
+func (sm *systemdManager) getServiceDetails(string) (systemd.ServiceDetails, error) {
+	return nil, errors.New("systemd manager unavailable")
+}

agent/systemd_nonlinux_test.go

@@ -0,0 +1,53 @@
+//go:build !linux && testing
+
+package agent
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestNewSystemdManager(t *testing.T) {
+	manager, err := newSystemdManager()
+	assert.NoError(t, err)
+	assert.NotNil(t, manager)
+}
+
+func TestSystemdManagerGetServiceStats(t *testing.T) {
+	manager, err := newSystemdManager()
+	assert.NoError(t, err)
+
+	// Test with refresh = true
+	result := manager.getServiceStats(true)
+	assert.Nil(t, result)
+
+	// Test with refresh = false
+	result = manager.getServiceStats(false)
+	assert.Nil(t, result)
+}
+
+func TestSystemdManagerGetServiceDetails(t *testing.T) {
+	manager, err := newSystemdManager()
+	assert.NoError(t, err)
+
+	result, err := manager.getServiceDetails("any-service")
+	assert.Error(t, err)
+	assert.Equal(t, "systemd manager unavailable", err.Error())
+	assert.Nil(t, result)
+
+	// Test with empty service name
+	result, err = manager.getServiceDetails("")
+	assert.Error(t, err)
+	assert.Equal(t, "systemd manager unavailable", err.Error())
+	assert.Nil(t, result)
+}
+
+func TestSystemdManagerFields(t *testing.T) {
+	manager, err := newSystemdManager()
+	assert.NoError(t, err)
+
+	// The non-linux manager should be a simple struct with no special fields
+	// We can't test private fields directly, but we can test the methods work
+	assert.NotNil(t, manager)
+}

agent/systemd_test.go

@@ -0,0 +1,188 @@
+//go:build linux && testing
+
+package agent
+
+import (
+	"os"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestUnescapeServiceName(t *testing.T) {
+	tests := []struct {
+		input    string
+		expected string
+	}{
+		{"nginx.service", "nginx.service"},                                     // No escaping needed
+		{"test\\x2dwith\\x2ddashes.service", "test-with-dashes.service"},       // \x2d is dash
+		{"service\\x20with\\x20spaces.service", "service with spaces.service"}, // \x20 is space
+		{"mixed\\x2dand\\x2dnormal", "mixed-and-normal"},                       // Mixed escaped and normal
+		{"no-escape-here", "no-escape-here"},                                   // No escape sequences
+		{"", ""},                                                               // Empty string
+		{"\\x2d\\x2d", "--"},                                                   // Multiple escapes
+	}
+
+	for _, test := range tests {
+		t.Run(test.input, func(t *testing.T) {
+			result := unescapeServiceName(test.input)
+			assert.Equal(t, test.expected, result)
+		})
+	}
+}
+
+func TestUnescapeServiceNameInvalid(t *testing.T) {
+	// Test invalid escape sequences - should return original string
+	invalidInputs := []string{
+		"invalid\\x",   // Incomplete escape
+		"invalid\\xZZ", // Invalid hex
+		"invalid\\x2",  // Incomplete hex
+		"invalid\\xyz", // Not a valid escape
+	}
+
+	for _, input := range invalidInputs {
+		t.Run(input, func(t *testing.T) {
+			result := unescapeServiceName(input)
+			assert.Equal(t, input, result, "Invalid escape sequences should return original string")
+		})
+	}
+}
+
+func TestIsSystemdAvailable(t *testing.T) {
+	// Note: This test's result will vary based on the actual system running the tests
+	// On systems with systemd, it should return true
+	// On systems without systemd, it should return false
+	result := isSystemdAvailable()
+
+	// Check if either the /run/systemd/system directory exists or PID 1 is systemd
+	runSystemdExists := false
+	if _, err := os.Stat("/run/systemd/system"); err == nil {
+		runSystemdExists = true
+	}
+
+	pid1IsSystemd := false
+	if data, err := os.ReadFile("/proc/1/comm"); err == nil {
+		pid1IsSystemd = strings.TrimSpace(string(data)) == "systemd"
+	}
+
+	expected := runSystemdExists || pid1IsSystemd
+
+	assert.Equal(t, expected, result, "isSystemdAvailable should correctly detect systemd presence")
+
+	// Log the result for informational purposes
+	if result {
+		t.Log("Systemd is available on this system")
+	} else {
+		t.Log("Systemd is not available on this system")
+	}
+}
+
+func TestGetServicePatterns(t *testing.T) {
+	tests := []struct {
+		name           string
+		prefixedEnv    string
+		unprefixedEnv  string
+		expected       []string
+		cleanupEnvVars bool
+	}{
+		{
+			name:           "default when no env var set",
+			prefixedEnv:    "",
+			unprefixedEnv:  "",
+			expected:       []string{"*.service"},
+			cleanupEnvVars: true,
+		},
+		{
+			name:           "single pattern with prefixed env",
+			prefixedEnv:    "nginx",
+			unprefixedEnv:  "",
+			expected:       []string{"nginx.service"},
+			cleanupEnvVars: true,
+		},
+		{
+			name:           "single pattern with unprefixed env",
+			prefixedEnv:    "",
+			unprefixedEnv:  "nginx",
+			expected:       []string{"nginx.service"},
+			cleanupEnvVars: true,
+		},
+		{
+			name:           "prefixed env takes precedence",
+			prefixedEnv:    "nginx",
+			unprefixedEnv:  "apache",
+			expected:       []string{"nginx.service"},
+			cleanupEnvVars: true,
+		},
+		{
+			name:           "multiple patterns",
+			prefixedEnv:    "nginx,apache,postgresql",
+			unprefixedEnv:  "",
+			expected:       []string{"nginx.service", "apache.service", "postgresql.service"},
+			cleanupEnvVars: true,
+		},
+		{
+			name:           "patterns with .service suffix",
+			prefixedEnv:    "nginx.service,apache.service",
+			unprefixedEnv:  "",
+			expected:       []string{"nginx.service", "apache.service"},
+			cleanupEnvVars: true,
+		},
+		{
+			name:           "mixed patterns with and without suffix",
+			prefixedEnv:    "nginx.service,apache,postgresql.service",
+			unprefixedEnv:  "",
+			expected:       []string{"nginx.service", "apache.service", "postgresql.service"},
+			cleanupEnvVars: true,
+		},
+		{
+			name:           "patterns with whitespace",
+			prefixedEnv:    " nginx , apache , postgresql ",
+			unprefixedEnv:  "",
+			expected:       []string{"nginx.service", "apache.service", "postgresql.service"},
+			cleanupEnvVars: true,
+		},
+		{
+			name:           "empty patterns are skipped",
+			prefixedEnv:    "nginx,,apache,  ,postgresql",
+			unprefixedEnv:  "",
+			expected:       []string{"nginx.service", "apache.service", "postgresql.service"},
+			cleanupEnvVars: true,
+		},
+		{
+			name:           "wildcard pattern",
+			prefixedEnv:    "*nginx*,*apache*",
+			unprefixedEnv:  "",
+			expected:       []string{"*nginx*.service", "*apache*.service"},
+			cleanupEnvVars: true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			// Clean up any existing env vars
+			os.Unsetenv("BESZEL_AGENT_SERVICE_PATTERNS")
+			os.Unsetenv("SERVICE_PATTERNS")
+
+			// Set up environment variables
+			if tt.prefixedEnv != "" {
+				os.Setenv("BESZEL_AGENT_SERVICE_PATTERNS", tt.prefixedEnv)
+			}
+			if tt.unprefixedEnv != "" {
+				os.Setenv("SERVICE_PATTERNS", tt.unprefixedEnv)
+			}
+
+			// Run the function
+			result := getServicePatterns()
+
+			// Verify results
+			assert.Equal(t, tt.expected, result, "Patterns should match expected values")
+
+			// Cleanup
+			if tt.cleanupEnvVars {
+				os.Unsetenv("BESZEL_AGENT_SERVICE_PATTERNS")
+				os.Unsetenv("SERVICE_PATTERNS")
+			}
+		})
+	}
+}

agent/test-data/container.json

@@ -0,0 +1,24 @@
+{
+	"cpu_stats": {
+		"cpu_usage": {
+			"total_usage": 312055276000
+		},
+		"system_cpu_usage": 1366399830000000
+	},
+	"memory_stats": {
+		"usage": 507400192,
+		"stats": {
+			"inactive_file": 165130240
+		}
+	},
+	"networks": {
+		"eth0": {
+			"tx_bytes": 20376558,
+			"rx_bytes": 537029455
+		},
+		"eth1": {
+			"tx_bytes": 2003766,
+			"rx_bytes": 6241
+		}
+	}
+}

agent/test-data/container2.json

@@ -0,0 +1,24 @@
+{
+	"cpu_stats": {
+		"cpu_usage": {
+			"total_usage": 314891801000
+		},
+		"system_cpu_usage": 1368474900000000
+	},
+	"memory_stats": {
+		"usage": 507400192,
+		"stats": {
+			"inactive_file": 165130240
+		}
+	},
+	"networks": {
+		"eth0": {
+			"tx_bytes": 20376558,
+			"rx_bytes": 537029455
+		},
+		"eth1": {
+			"tx_bytes": 2003766,
+			"rx_bytes": 6241
+		}
+	}
+}

agent/test-data/smart/nvme0.json

@@ -0,0 +1,272 @@
+{
+  "json_format_version": [
+    1,
+    0
+  ],
+  "smartctl": {
+    "version": [
+      7,
+      5
+    ],
+    "pre_release": false,
+    "svn_revision": "5714",
+    "platform_info": "x86_64-linux-6.17.1-2-cachyos",
+    "build_info": "(local build)",
+    "argv": [
+      "smartctl",
+      "-aj",
+      "/dev/nvme0"
+    ],
+    "exit_status": 0
+  },
+  "local_time": {
+    "time_t": 1761507494,
+    "asctime": "Sun Oct 26 15:38:14 2025 EDT"
+  },
+  "device": {
+    "name": "/dev/nvme0",
+    "info_name": "/dev/nvme0",
+    "type": "nvme",
+    "protocol": "NVMe"
+  },
+  "model_name": "PELADN 512GB",
+  "serial_number": "2024031600129",
+  "firmware_version": "VC2S038E",
+  "nvme_pci_vendor": {
+    "id": 4332,
+    "subsystem_id": 4332
+  },
+  "nvme_ieee_oui_identifier": 57420,
+  "nvme_controller_id": 1,
+  "nvme_version": {
+    "string": "1.4",
+    "value": 66560
+  },
+  "nvme_number_of_namespaces": 1,
+  "nvme_namespaces": [
+    {
+      "id": 1,
+      "size": {
+        "blocks": 1000215216,
+        "bytes": 512110190592
+      },
+      "capacity": {
+        "blocks": 1000215216,
+        "bytes": 512110190592
+      },
+      "utilization": {
+        "blocks": 1000215216,
+        "bytes": 512110190592
+      },
+      "formatted_lba_size": 512,
+      "eui64": {
+        "oui": 57420,
+        "ext_id": 112094110470
+      },
+      "features": {
+        "value": 0,
+        "thin_provisioning": false,
+        "na_fields": false,
+        "dealloc_or_unwritten_block_error": false,
+        "uid_reuse": false,
+        "np_fields": false,
+        "other": 0
+      },
+      "lba_formats": [
+        {
+          "formatted": true,
+          "data_bytes": 512,
+          "metadata_bytes": 0,
+          "relative_performance": 0
+        }
+      ]
+    }
+  ],
+  "user_capacity": {
+    "blocks": 1000215216,
+    "bytes": 512110190592
+  },
+  "logical_block_size": 512,
+  "smart_support": {
+    "available": true,
+    "enabled": true
+  },
+  "nvme_firmware_update_capabilities": {
+    "value": 2,
+    "slots": 1,
+    "first_slot_is_read_only": false,
+    "activiation_without_reset": false,
+    "multiple_update_detection": false,
+    "other": 0
+  },
+  "nvme_optional_admin_commands": {
+    "value": 23,
+    "security_send_receive": true,
+    "format_nvm": true,
+    "firmware_download": true,
+    "namespace_management": false,
+    "self_test": true,
+    "directives": false,
+    "mi_send_receive": false,
+    "virtualization_management": false,
+    "doorbell_buffer_config": false,
+    "get_lba_status": false,
+    "command_and_feature_lockdown": false,
+    "other": 0
+  },
+  "nvme_optional_nvm_commands": {
+    "value": 94,
+    "compare": false,
+    "write_uncorrectable": true,
+    "dataset_management": true,
+    "write_zeroes": true,
+    "save_select_feature_nonzero": true,
+    "reservations": false,
+    "timestamp": true,
+    "verify": false,
+    "copy": false,
+    "other": 0
+  },
+  "nvme_log_page_attributes": {
+    "value": 2,
+    "smart_health_per_namespace": false,
+    "commands_effects_log": true,
+    "extended_get_log_page_cmd": false,
+    "telemetry_log": false,
+    "persistent_event_log": false,
+    "supported_log_pages_log": false,
+    "telemetry_data_area_4": false,
+    "other": 0
+  },
+  "nvme_maximum_data_transfer_pages": 32,
+  "nvme_composite_temperature_threshold": {
+    "warning": 100,
+    "critical": 110
+  },
+  "temperature": {
+    "op_limit_max": 100,
+    "critical_limit_max": 110,
+    "current": 61
+  },
+  "nvme_power_states": [
+    {
+      "non_operational_state": false,
+      "relative_read_latency": 0,
+      "relative_read_throughput": 0,
+      "relative_write_latency": 0,
+      "relative_write_throughput": 0,
+      "entry_latency_us": 230000,
+      "exit_latency_us": 50000,
+      "max_power": {
+        "value": 800,
+        "scale": 2,
+        "units_per_watt": 100
+      }
+    },
+    {
+      "non_operational_state": false,
+      "relative_read_latency": 1,
+      "relative_read_throughput": 1,
+      "relative_write_latency": 1,
+      "relative_write_throughput": 1,
+      "entry_latency_us": 4000,
+      "exit_latency_us": 50000,
+      "max_power": {
+        "value": 400,
+        "scale": 2,
+        "units_per_watt": 100
+      }
+    },
+    {
+      "non_operational_state": false,
+      "relative_read_latency": 2,
+      "relative_read_throughput": 2,
+      "relative_write_latency": 2,
+      "relative_write_throughput": 2,
+      "entry_latency_us": 4000,
+      "exit_latency_us": 250000,
+      "max_power": {
+        "value": 300,
+        "scale": 2,
+        "units_per_watt": 100
+      }
+    },
+    {
+      "non_operational_state": true,
+      "relative_read_latency": 3,
+      "relative_read_throughput": 3,
+      "relative_write_latency": 3,
+      "relative_write_throughput": 3,
+      "entry_latency_us": 5000,
+      "exit_latency_us": 10000,
+      "max_power": {
+        "value": 300,
+        "scale": 1,
+        "units_per_watt": 10000
+      }
+    },
+    {
+      "non_operational_state": true,
+      "relative_read_latency": 4,
+      "relative_read_throughput": 4,
+      "relative_write_latency": 4,
+      "relative_write_throughput": 4,
+      "entry_latency_us": 54000,
+      "exit_latency_us": 45000,
+      "max_power": {
+        "value": 50,
+        "scale": 1,
+        "units_per_watt": 10000
+      }
+    }
+  ],
+  "smart_status": {
+    "passed": true,
+    "nvme": {
+      "value": 0
+    }
+  },
+  "nvme_smart_health_information_log": {
+    "nsid": -1,
+    "critical_warning": 0,
+    "temperature": 61,
+    "available_spare": 100,
+    "available_spare_threshold": 32,
+    "percentage_used": 0,
+    "data_units_read": 6573104,
+    "data_units_written": 16040567,
+    "host_reads": 63241130,
+    "host_writes": 253050006,
+    "controller_busy_time": 0,
+    "power_cycles": 430,
+    "power_on_hours": 4399,
+    "unsafe_shutdowns": 44,
+    "media_errors": 0,
+    "num_err_log_entries": 0,
+    "warning_temp_time": 0,
+    "critical_comp_time": 0
+  },
+  "spare_available": {
+    "current_percent": 100,
+    "threshold_percent": 32
+  },
+  "endurance_used": {
+    "current_percent": 0
+  },
+  "power_cycle_count": 430,
+  "power_on_time": {
+    "hours": 4399
+  },
+  "nvme_error_information_log": {
+    "size": 8,
+    "read": 8,
+    "unread": 0
+  },
+  "nvme_self_test_log": {
+    "nsid": -1,
+    "current_self_test_operation": {
+      "value": 0,
+      "string": "No self-test in progress"
+    }
+  }
+}

agent/test-data/smart/scan.json

@@ -0,0 +1,36 @@
+{
+  "json_format_version": [
+    1,
+    0
+  ],
+  "smartctl": {
+    "version": [
+      7,
+      5
+    ],
+    "pre_release": false,
+    "svn_revision": "5714",
+    "platform_info": "x86_64-linux-6.17.1-2-cachyos",
+    "build_info": "(local build)",
+    "argv": [
+      "smartctl",
+      "--scan",
+      "-j"
+    ],
+    "exit_status": 0
+  },
+  "devices": [
+    {
+      "name": "/dev/sda",
+      "info_name": "/dev/sda [SAT]",
+      "type": "sat",
+      "protocol": "ATA"
+    },
+    {
+      "name": "/dev/nvme0",
+      "info_name": "/dev/nvme0",
+      "type": "nvme",
+      "protocol": "NVMe"
+    }
+  ]
+}

agent/test-data/smart/scsi.json

@@ -0,0 +1,125 @@
+{
+   "json_format_version": [
+      1,
+      0
+   ],
+   "smartctl": {
+      "version": [
+         7,
+         3
+      ],
+      "svn_revision": "5338",
+      "platform_info": "x86_64-linux-6.12.43+deb12-amd64",
+      "build_info": "(local build)",
+      "argv": [
+         "smartctl",
+         "-aj",
+         "/dev/sde"
+      ],
+      "exit_status": 0
+   },
+   "local_time": {
+      "time_t": 1761502142,
+      "asctime": "Sun Oct 21 21:09:02 2025 MSK"
+   },
+   "device": {
+      "name": "/dev/sde",
+      "info_name": "/dev/sde",
+      "type": "scsi",
+      "protocol": "SCSI"
+   },
+   "scsi_vendor": "YADRO",
+   "scsi_product": "WUH721414AL4204",
+   "scsi_model_name": "YADRO WUH721414AL4204",
+   "scsi_revision": "C240",
+   "scsi_version": "SPC-4",
+   "user_capacity": {
+      "blocks": 3418095616,
+      "bytes": 14000519643136
+   },
+   "logical_block_size": 4096,
+   "scsi_lb_provisioning": {
+      "name": "fully provisioned",
+      "value": 0,
+      "management_enabled": {
+         "name": "LBPME",
+         "value": 0
+      },
+      "read_zeros": {
+         "name": "LBPRZ",
+         "value": 0
+      }
+   },
+   "rotation_rate": 7200,
+   "form_factor": {
+      "scsi_value": 2,
+      "name": "3.5 inches"
+   },
+   "logical_unit_id": "0x5000cca29063dc00",
+   "serial_number": "9YHSDH9B",
+   "device_type": {
+      "scsi_terminology": "Peripheral Device Type [PDT]",
+      "scsi_value": 0,
+      "name": "disk"
+   },
+   "scsi_transport_protocol": {
+      "name": "SAS (SPL-4)",
+      "value": 6
+   },
+   "smart_support": {
+      "available": true,
+      "enabled": true
+   },
+   "temperature_warning": {
+      "enabled": true
+   },
+   "smart_status": {
+      "passed": true
+   },
+   "temperature": {
+      "current": 34,
+      "drive_trip": 85
+   },
+   "power_on_time": {
+      "hours": 458,
+      "minutes": 25
+   },
+   "scsi_start_stop_cycle_counter": {
+      "year_of_manufacture": "2022",
+      "week_of_manufacture": "41",
+      "specified_cycle_count_over_device_lifetime": 50000,
+      "accumulated_start_stop_cycles": 2,
+      "specified_load_unload_count_over_device_lifetime": 600000,
+      "accumulated_load_unload_cycles": 418
+   },
+   "scsi_grown_defect_list": 0,
+   "scsi_error_counter_log": {
+      "read": {
+         "errors_corrected_by_eccfast": 0,
+         "errors_corrected_by_eccdelayed": 0,
+         "errors_corrected_by_rereads_rewrites": 0,
+         "total_errors_corrected": 0,
+         "correction_algorithm_invocations": 346,
+         "gigabytes_processed": "3,641",
+         "total_uncorrected_errors": 0
+      },
+      "write": {
+         "errors_corrected_by_eccfast": 0,
+         "errors_corrected_by_eccdelayed": 0,
+         "errors_corrected_by_rereads_rewrites": 0,
+         "total_errors_corrected": 0,
+         "correction_algorithm_invocations": 4052,
+         "gigabytes_processed": "2124,590",
+         "total_uncorrected_errors": 0
+      },
+      "verify": {
+         "errors_corrected_by_eccfast": 0,
+         "errors_corrected_by_eccdelayed": 0,
+         "errors_corrected_by_rereads_rewrites": 0,
+         "total_errors_corrected": 0,
+         "correction_algorithm_invocations": 223,
+         "gigabytes_processed": "0,000",
+         "total_uncorrected_errors": 0
+      }
+   }
+}

agent/test-data/system_info.json

@@ -0,0 +1,17 @@
+{
+  "ID": "7TRN:IPZB:QYBB:VPBQ:UMPP:KARE:6ZNR:XE6T:7EWV:PKF4:ZOJD:TPYS",
+  "Containers": 14,
+  "ContainersRunning": 3,
+  "ContainersPaused": 1,
+  "ContainersStopped": 10,
+  "Images": 508,
+  "Driver": "overlay2",
+  "KernelVersion": "6.8.0-31-generic",
+  "OperatingSystem": "Ubuntu 24.04 LTS",
+  "OSVersion": "24.04",
+  "OSType": "linux",
+  "Architecture": "x86_64",
+  "NCPU": 4,
+  "MemTotal": 2095882240,
+  "ServerVersion": "27.0.1"
+}

agent/tools/fetchsmartctl/main.go

@@ -0,0 +1,130 @@
+package main
+
+import (
+	"crypto/sha1"
+	"crypto/sha256"
+	"encoding/hex"
+	"flag"
+	"fmt"
+	"hash"
+	"io"
+	"net/http"
+	"os"
+	"path/filepath"
+	"strings"
+	"time"
+)
+
+// Download smartctl.exe from the given URL and save it to the given destination.
+// This is used to embed smartctl.exe in the Windows build.
+
+func main() {
+	url := flag.String("url", "", "URL to download smartctl.exe from (required)")
+	out := flag.String("out", "", "Destination path for smartctl.exe (required)")
+	sha := flag.String("sha", "", "Optional SHA1/SHA256 checksum for integrity validation")
+	force := flag.Bool("force", false, "Force re-download even if destination exists")
+	flag.Parse()
+
+	if *url == "" || *out == "" {
+		fatalf("-url and -out are required")
+	}
+
+	if !*force {
+		if info, err := os.Stat(*out); err == nil && info.Size() > 0 {
+			fmt.Println("smartctl.exe already present, skipping download")
+			return
+		}
+	}
+
+	if err := downloadFile(*url, *out, *sha); err != nil {
+		fatalf("download failed: %v", err)
+	}
+}
+
+func downloadFile(url, dest, shaHex string) error {
+	// Prepare destination
+	if err := os.MkdirAll(filepath.Dir(dest), 0o755); err != nil {
+		return fmt.Errorf("create dir: %w", err)
+	}
+
+	// HTTP client
+	client := &http.Client{Timeout: 60 * time.Second}
+	req, err := http.NewRequest(http.MethodGet, url, nil)
+	if err != nil {
+		return fmt.Errorf("new request: %w", err)
+	}
+	req.Header.Set("User-Agent", "beszel-fetchsmartctl/1.0")
+
+	resp, err := client.Do(req)
+	if err != nil {
+		return fmt.Errorf("http get: %w", err)
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode < 200 || resp.StatusCode >= 300 {
+		return fmt.Errorf("unexpected HTTP status: %s", resp.Status)
+	}
+
+	tmp := dest + ".tmp"
+	f, err := os.OpenFile(tmp, os.O_CREATE|os.O_TRUNC|os.O_WRONLY, 0o644)
+	if err != nil {
+		return fmt.Errorf("open tmp: %w", err)
+	}
+
+	// Determine hash algorithm based on length (SHA1=40, SHA256=64)
+	var hasher hash.Hash
+	if shaHex := strings.TrimSpace(shaHex); shaHex != "" {
+		cleanSha := strings.ToLower(strings.ReplaceAll(shaHex, " ", ""))
+		switch len(cleanSha) {
+		case 40:
+			hasher = sha1.New()
+		case 64:
+			hasher = sha256.New()
+		default:
+			f.Close()
+			os.Remove(tmp)
+			return fmt.Errorf("unsupported hash length: %d (expected 40 for SHA1 or 64 for SHA256)", len(cleanSha))
+		}
+	}
+
+	var mw io.Writer = f
+	if hasher != nil {
+		mw = io.MultiWriter(f, hasher)
+	}
+	if _, err := io.Copy(mw, resp.Body); err != nil {
+		f.Close()
+		os.Remove(tmp)
+		return fmt.Errorf("write tmp: %w", err)
+	}
+	if err := f.Close(); err != nil {
+		os.Remove(tmp)
+		return fmt.Errorf("close tmp: %w", err)
+	}
+
+	if hasher != nil && shaHex != "" {
+		cleanSha := strings.ToLower(strings.ReplaceAll(strings.TrimSpace(shaHex), " ", ""))
+		got := strings.ToLower(hex.EncodeToString(hasher.Sum(nil)))
+		if got != cleanSha {
+			os.Remove(tmp)
+			return fmt.Errorf("hash mismatch: got %s want %s", got, cleanSha)
+		}
+	}
+
+	// Make executable and move into place
+	if err := os.Chmod(tmp, 0o755); err != nil {
+		os.Remove(tmp)
+		return fmt.Errorf("chmod: %w", err)
+	}
+	if err := os.Rename(tmp, dest); err != nil {
+		os.Remove(tmp)
+		return fmt.Errorf("rename: %w", err)
+	}
+
+	fmt.Println("smartctl.exe downloaded to", dest)
+	return nil
+}
+
+func fatalf(format string, a ...any) {
+	fmt.Fprintf(os.Stderr, format+"\n", a...)
+	os.Exit(1)
+}

agent/update.go

@@ -1,12 +1,14 @@
 package agent
 
 import (
-	"beszel/internal/ghupdate"
 	"fmt"
 	"log"
 	"os"
 	"os/exec"
+	"runtime"
 	"strings"
+
+	"github.com/henrygd/beszel/internal/ghupdate"
 )
 
 // restarter knows how to restart the beszel-agent service.
@@ -28,21 +30,32 @@ func (s *systemdRestarter) Restart() error {
 type openRCRestarter struct{ cmd string }
 
 func (o *openRCRestarter) Restart() error {
-	if err := exec.Command(o.cmd, "status", "beszel-agent").Run(); err != nil {
+	if err := exec.Command(o.cmd, "beszel-agent", "status").Run(); err != nil {
 		return nil
 	}
 	ghupdate.ColorPrint(ghupdate.ColorYellow, "Restarting beszel-agent via OpenRC…")
-	return exec.Command(o.cmd, "restart", "beszel-agent").Run()
+	return exec.Command(o.cmd, "beszel-agent", "restart").Run()
 }
 
 type openWRTRestarter struct{ cmd string }
 
 func (w *openWRTRestarter) Restart() error {
-	if err := exec.Command(w.cmd, "running", "beszel-agent").Run(); err != nil {
+	// https://openwrt.org/docs/guide-user/base-system/managing_services?s[]=service
+	if err := exec.Command("/etc/init.d/beszel-agent", "running").Run(); err != nil {
 		return nil
 	}
 	ghupdate.ColorPrint(ghupdate.ColorYellow, "Restarting beszel-agent via procd…")
-	return exec.Command(w.cmd, "restart", "beszel-agent").Run()
+	return exec.Command("/etc/init.d/beszel-agent", "restart").Run()
+}
+
+type freeBSDRestarter struct{ cmd string }
+
+func (f *freeBSDRestarter) Restart() error {
+	if err := exec.Command(f.cmd, "beszel-agent", "status").Run(); err != nil {
+		return nil
+	}
+	ghupdate.ColorPrint(ghupdate.ColorYellow, "Restarting beszel-agent via FreeBSD rc…")
+	return exec.Command(f.cmd, "beszel-agent", "restart").Run()
 }
 
 func detectRestarter() restarter {
@@ -52,8 +65,13 @@ func detectRestarter() restarter {
 	if path, err := exec.LookPath("rc-service"); err == nil {
 		return &openRCRestarter{cmd: path}
 	}
+    if path, err := exec.LookPath("procd"); err == nil {
+        return &openWRTRestarter{cmd: path}
+    }
 	if path, err := exec.LookPath("service"); err == nil {
-		return &openWRTRestarter{cmd: path}
+		if runtime.GOOS == "freebsd" {
+			return &freeBSDRestarter{cmd: path}
+		}
 	}
 	return nil
 }

beszel.go

@@ -0,0 +1,18 @@
+// Package beszel provides core application constants and version information
+// which are used throughout the application.
+package beszel
+
+import "github.com/blang/semver"
+
+const (
+	// Version is the current version of the application.
+	Version = "0.18.0"
+	// AppName is the name of the application.
+	AppName = "beszel"
+)
+
+// MinVersionCbor is the minimum supported version for CBOR compatibility.
+var MinVersionCbor = semver.MustParse("0.12.0")
+
+// MinVersionAgentResponse is the minimum supported version for AgentResponse compatibility.
+var MinVersionAgentResponse = semver.MustParse("0.13.0")

beszel/Makefile

@@ -1,98 +0,0 @@
-# Default OS/ARCH values
-OS ?= $(shell go env GOOS)
-ARCH ?= $(shell go env GOARCH)
-# Skip building the web UI if true
-SKIP_WEB ?= false
-
-# Set executable extension based on target OS
-EXE_EXT := $(if $(filter windows,$(OS)),.exe,)
-
-.PHONY: tidy build-agent build-hub build clean lint dev-server dev-agent dev-hub dev generate-locales
-.DEFAULT_GOAL := build
-
-clean:
-	go clean
-	rm -rf ./build
-
-lint:
-	golangci-lint run
-
-test: export GOEXPERIMENT=synctest
-test:
-	go test -tags=testing ./...
-
-tidy:
-	go mod tidy
-
-build-web-ui:
-	@if command -v bun >/dev/null 2>&1; then \
-		bun install --cwd ./site && \
-		bun run --cwd ./site build; \
-	else \
-		npm install --prefix ./site && \
-		npm run --prefix ./site build; \
-	fi
-
-# Conditional .NET build - only for Windows
-build-dotnet-conditional:
-	@if [ "$(OS)" = "windows" ]; then \
-		echo "Building .NET executable for Windows..."; \
-		if command -v dotnet >/dev/null 2>&1; then \
-			rm -rf ./internal/agent/lhm/bin; \
-			dotnet build -c Release ./internal/agent/lhm/beszel_lhm.csproj; \
-		else \
-			echo "Error: dotnet not found. Install .NET SDK to build Windows agent."; \
-			exit 1; \
-		fi; \
-	fi
-
-# Update build-agent to include conditional .NET build
-build-agent: tidy build-dotnet-conditional
-	GOOS=$(OS) GOARCH=$(ARCH) go build -o ./build/beszel-agent_$(OS)_$(ARCH)$(EXE_EXT) -ldflags "-w -s" beszel/cmd/agent
-
-build-hub: tidy $(if $(filter false,$(SKIP_WEB)),build-web-ui)
-	GOOS=$(OS) GOARCH=$(ARCH) go build -o ./build/beszel_$(OS)_$(ARCH)$(EXE_EXT) -ldflags "-w -s" beszel/cmd/hub
-
-build: build-agent build-hub
-
-generate-locales:
-	@if [ ! -f ./site/src/locales/en/en.ts ]; then \
-		echo "Generating locales..."; \
-		command -v bun >/dev/null 2>&1 && cd ./site && bun install && bun run sync || cd ./site && npm install && npm run sync; \
-	fi
-
-dev-server: generate-locales
-	cd ./site
-	@if command -v bun >/dev/null 2>&1; then \
-		cd ./site && bun run dev --host 0.0.0.0; \
-	else \
-		cd ./site && npm run dev --host 0.0.0.0; \
-	fi
-
-dev-hub: export ENV=dev
-dev-hub:
-	mkdir -p ./site/dist && touch ./site/dist/index.html
-	@if command -v entr >/dev/null 2>&1; then \
-		find ./cmd/hub/*.go ./internal/{alerts,hub,records,users}/*.go | entr -r -s "cd ./cmd/hub && go run . serve --http 0.0.0.0:8090"; \
-	else \
-		cd ./cmd/hub && go run . serve --http 0.0.0.0:8090; \
-	fi
-
-dev-agent:
-	@if command -v entr >/dev/null 2>&1; then \
-		find ./cmd/agent/*.go ./internal/agent/*.go | entr -r go run beszel/cmd/agent; \
-	else \
-		go run beszel/cmd/agent; \
-	fi
-	
-build-dotnet:
-	@if command -v dotnet >/dev/null 2>&1; then \
-		rm -rf ./internal/agent/lhm/bin; \
-		dotnet build -c Release ./internal/agent/lhm/beszel_lhm.csproj; \
-	else \
-		echo "dotnet not found"; \
-	fi
-
-
-# KEY="..." make -j dev
-dev: dev-server dev-hub dev-agent

beszel/dockerfile_agent_nvidia

@@ -1,21 +0,0 @@
-FROM --platform=$BUILDPLATFORM golang:alpine AS builder
-
-WORKDIR /app
-
-COPY go.mod go.sum ./
-# RUN go mod download
-COPY *.go ./
-COPY cmd ./cmd
-COPY internal ./internal
-
-# Build
-ARG TARGETOS TARGETARCH
-RUN CGO_ENABLED=0 GOGC=75 GOOS=$TARGETOS GOARCH=$TARGETARCH go build -ldflags "-w -s" -o /agent ./cmd/agent
-
-# --------------------------
-# Final image: GPU-enabled agent with nvidia-smi
-# --------------------------
-FROM nvidia/cuda:12.2.2-base-ubuntu22.04
-COPY --from=builder /agent /agent
-
-ENTRYPOINT ["/agent"]

beszel/internal/agent/agent.go

@@ -1,182 +0,0 @@
-// Package agent handles the agent's SSH server and system stats collection.
-package agent
-
-import (
-	"beszel"
-	"beszel/internal/entities/system"
-	"crypto/sha256"
-	"encoding/hex"
-	"log/slog"
-	"os"
-	"path/filepath"
-	"strings"
-	"sync"
-	"time"
-
-	"github.com/gliderlabs/ssh"
-	"github.com/shirou/gopsutil/v4/host"
-	gossh "golang.org/x/crypto/ssh"
-)
-
-type Agent struct {
-	sync.Mutex                                   // Used to lock agent while collecting data
-	debug             bool                       // true if LOG_LEVEL is set to debug
-	zfs               bool                       // true if system has arcstats
-	memCalc           string                     // Memory calculation formula
-	fsNames           []string                   // List of filesystem device names being monitored
-	fsStats           map[string]*system.FsStats // Keeps track of disk stats for each filesystem
-	netInterfaces     map[string]struct{}        // Stores all valid network interfaces
-	netIoStats        system.NetIoStats          // Keeps track of bandwidth usage
-	dockerManager     *dockerManager             // Manages Docker API requests
-	sensorConfig      *SensorConfig              // Sensors config
-	systemInfo        system.Info                // Host system info
-	gpuManager        *GPUManager                // Manages GPU data
-	cache             *SessionCache              // Cache for system stats based on primary session ID
-	connectionManager *ConnectionManager         // Channel to signal connection events
-	server            *ssh.Server                // SSH server
-	dataDir           string                     // Directory for persisting data
-	keys              []gossh.PublicKey          // SSH public keys
-}
-
-// NewAgent creates a new agent with the given data directory for persisting data.
-// If the data directory is not set, it will attempt to find the optimal directory.
-func NewAgent(dataDir ...string) (agent *Agent, err error) {
-	agent = &Agent{
-		fsStats: make(map[string]*system.FsStats),
-		cache:   NewSessionCache(69 * time.Second),
-	}
-
-	agent.dataDir, err = getDataDir(dataDir...)
-	if err != nil {
-		slog.Warn("Data directory not found")
-	} else {
-		slog.Info("Data directory", "path", agent.dataDir)
-	}
-
-	agent.memCalc, _ = GetEnv("MEM_CALC")
-	agent.sensorConfig = agent.newSensorConfig()
-	// Set up slog with a log level determined by the LOG_LEVEL env var
-	if logLevelStr, exists := GetEnv("LOG_LEVEL"); exists {
-		switch strings.ToLower(logLevelStr) {
-		case "debug":
-			agent.debug = true
-			slog.SetLogLoggerLevel(slog.LevelDebug)
-		case "warn":
-			slog.SetLogLoggerLevel(slog.LevelWarn)
-		case "error":
-			slog.SetLogLoggerLevel(slog.LevelError)
-		}
-	}
-
-	slog.Debug(beszel.Version)
-
-	// initialize system info
-	agent.initializeSystemInfo()
-
-	// initialize connection manager
-	agent.connectionManager = newConnectionManager(agent)
-
-	// initialize disk info
-	agent.initializeDiskInfo()
-
-	// initialize net io stats
-	agent.initializeNetIoStats()
-
-	// initialize docker manager
-	agent.dockerManager = newDockerManager(agent)
-
-	// initialize GPU manager
-	if gm, err := NewGPUManager(); err != nil {
-		slog.Debug("GPU", "err", err)
-	} else {
-		agent.gpuManager = gm
-	}
-
-	// if debugging, print stats
-	if agent.debug {
-		slog.Debug("Stats", "data", agent.gatherStats(""))
-	}
-
-	return agent, nil
-}
-
-// GetEnv retrieves an environment variable with a "BESZEL_AGENT_" prefix, or falls back to the unprefixed key.
-func GetEnv(key string) (value string, exists bool) {
-	if value, exists = os.LookupEnv("BESZEL_AGENT_" + key); exists {
-		return value, exists
-	}
-	// Fallback to the old unprefixed key
-	return os.LookupEnv(key)
-}
-
-func (a *Agent) gatherStats(sessionID string) *system.CombinedData {
-	a.Lock()
-	defer a.Unlock()
-
-	data, isCached := a.cache.Get(sessionID)
-	if isCached {
-		slog.Debug("Cached data", "session", sessionID)
-		return data
-	}
-
-	*data = system.CombinedData{
-		Stats: a.getSystemStats(),
-		Info:  a.systemInfo,
-	}
-	slog.Debug("System data", "data", data)
-
-	if a.dockerManager != nil {
-		if containerStats, err := a.dockerManager.getDockerStats(); err == nil {
-			data.Containers = containerStats
-			slog.Debug("Containers", "data", data.Containers)
-		} else {
-			slog.Debug("Containers", "err", err)
-		}
-	}
-
-	data.Stats.ExtraFs = make(map[string]*system.FsStats)
-	for name, stats := range a.fsStats {
-		if !stats.Root && stats.DiskTotal > 0 {
-			data.Stats.ExtraFs[name] = stats
-		}
-	}
-	slog.Debug("Extra FS", "data", data.Stats.ExtraFs)
-
-	a.cache.Set(sessionID, data)
-	return data
-}
-
-// StartAgent initializes and starts the agent with optional WebSocket connection
-func (a *Agent) Start(serverOptions ServerOptions) error {
-	a.keys = serverOptions.Keys
-	return a.connectionManager.Start(serverOptions)
-}
-
-func (a *Agent) getFingerprint() string {
-	// first look for a fingerprint in the data directory
-	if a.dataDir != "" {
-		if fp, err := os.ReadFile(filepath.Join(a.dataDir, "fingerprint")); err == nil {
-			return string(fp)
-		}
-	}
-
-	// if no fingerprint is found, generate one
-	fingerprint, err := host.HostID()
-	if err != nil || fingerprint == "" {
-		fingerprint = a.systemInfo.Hostname + a.systemInfo.CpuModel
-	}
-
-	// hash fingerprint
-	sum := sha256.Sum256([]byte(fingerprint))
-	fingerprint = hex.EncodeToString(sum[:24])
-
-	// save fingerprint to data directory
-	if a.dataDir != "" {
-		err = os.WriteFile(filepath.Join(a.dataDir, "fingerprint"), []byte(fingerprint), 0644)
-		if err != nil {
-			slog.Warn("Failed to save fingerprint", "err", err)
-		}
-	}
-
-	return fingerprint
-}

beszel/internal/agent/agent_cache.go

@@ -1,36 +0,0 @@
-package agent
-
-import (
-	"beszel/internal/entities/system"
-	"time"
-)
-
-// Not thread safe since we only access from gatherStats which is already locked
-type SessionCache struct {
-	data           *system.CombinedData
-	lastUpdate     time.Time
-	primarySession string
-	leaseTime      time.Duration
-}
-
-func NewSessionCache(leaseTime time.Duration) *SessionCache {
-	return &SessionCache{
-		leaseTime: leaseTime,
-		data:      &system.CombinedData{},
-	}
-}
-
-func (c *SessionCache) Get(sessionID string) (stats *system.CombinedData, isCached bool) {
-	if sessionID != c.primarySession && time.Since(c.lastUpdate) < c.leaseTime {
-		return c.data, true
-	}
-	return c.data, false
-}
-
-func (c *SessionCache) Set(sessionID string, data *system.CombinedData) {
-	if data != nil {
-		*c.data = *data
-	}
-	c.primarySession = sessionID
-	c.lastUpdate = time.Now()
-}

beszel/internal/agent/agent_cache_test.go

@@ -1,88 +0,0 @@
-//go:build testing
-// +build testing
-
-package agent
-
-import (
-	"beszel/internal/entities/system"
-	"testing"
-	"testing/synctest"
-	"time"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestSessionCache_GetSet(t *testing.T) {
-	synctest.Test(t, func(t *testing.T) {
-		cache := NewSessionCache(69 * time.Second)
-
-		testData := &system.CombinedData{
-			Info: system.Info{
-				Hostname: "test-host",
-				Cores:    4,
-			},
-			Stats: system.Stats{
-				Cpu:     50.0,
-				MemPct:  30.0,
-				DiskPct: 40.0,
-			},
-		}
-
-		// Test initial state - should not be cached
-		data, isCached := cache.Get("session1")
-		assert.False(t, isCached, "Expected no cached data initially")
-		assert.NotNil(t, data, "Expected data to be initialized")
-		// Set data for session1
-		cache.Set("session1", testData)
-
-		time.Sleep(15 * time.Second)
-
-		// Get data for a different session - should be cached
-		data, isCached = cache.Get("session2")
-		assert.True(t, isCached, "Expected data to be cached for non-primary session")
-		require.NotNil(t, data, "Expected cached data to be returned")
-		assert.Equal(t, "test-host", data.Info.Hostname, "Hostname should match test data")
-		assert.Equal(t, 4, data.Info.Cores, "Cores should match test data")
-		assert.Equal(t, 50.0, data.Stats.Cpu, "CPU should match test data")
-		assert.Equal(t, 30.0, data.Stats.MemPct, "Memory percentage should match test data")
-		assert.Equal(t, 40.0, data.Stats.DiskPct, "Disk percentage should match test data")
-
-		time.Sleep(10 * time.Second)
-
-		// Get data for the primary session - should not be cached
-		data, isCached = cache.Get("session1")
-		assert.False(t, isCached, "Expected data not to be cached for primary session")
-		require.NotNil(t, data, "Expected data to be returned even if not cached")
-		assert.Equal(t, "test-host", data.Info.Hostname, "Hostname should match test data")
-		// if not cached, agent will update the data
-		cache.Set("session1", testData)
-
-		time.Sleep(45 * time.Second)
-
-		// Get data for a different session - should still be cached
-		_, isCached = cache.Get("session2")
-		assert.True(t, isCached, "Expected data to be cached for non-primary session")
-
-		// Wait for the lease to expire
-		time.Sleep(30 * time.Second)
-
-		// Get data for session2 - should not be cached
-		_, isCached = cache.Get("session2")
-		assert.False(t, isCached, "Expected data not to be cached after lease expiration")
-	})
-}
-
-func TestSessionCache_NilData(t *testing.T) {
-	// Create a new SessionCache
-	cache := NewSessionCache(30 * time.Second)
-
-	// Test setting nil data (should not panic)
-	assert.NotPanics(t, func() {
-		cache.Set("session1", nil)
-	}, "Setting nil data should not panic")
-
-	// Get data - should not be nil even though we set nil
-	data, _ := cache.Get("session2")
-	assert.NotNil(t, data, "Expected data to not be nil after setting nil data")
-}

beszel/internal/agent/battery/battery.go

@@ -1,53 +0,0 @@
-//go:build !freebsd
-
-// Package battery provides functions to check if the system has a battery and to get the battery stats.
-package battery
-
-import (
-	"errors"
-	"log/slog"
-
-	"github.com/distatus/battery"
-)
-
-var systemHasBattery = false
-var haveCheckedBattery = false
-
-// HasReadableBattery checks if the system has a battery and returns true if it does.
-func HasReadableBattery() bool {
-	if haveCheckedBattery {
-		return systemHasBattery
-	}
-	haveCheckedBattery = true
-	bat, err := battery.Get(0)
-	if err == nil && bat != nil {
-		systemHasBattery = true
-	} else {
-		slog.Debug("No battery found", "err", err)
-	}
-	return systemHasBattery
-}
-
-// GetBatteryStats returns the current battery percent and charge state
-func GetBatteryStats() (batteryPercent uint8, batteryState uint8, err error) {
-	if !systemHasBattery {
-		return batteryPercent, batteryState, errors.ErrUnsupported
-	}
-	batteries, err := battery.GetAll()
-	if err != nil || len(batteries) == 0 {
-		return batteryPercent, batteryState, err
-	}
-	totalCapacity := float64(0)
-	totalCharge := float64(0)
-	for _, bat := range batteries {
-		if bat.Design != 0 {
-			totalCapacity += bat.Design
-		} else {
-			totalCapacity += bat.Full
-		}
-		totalCharge += bat.Current
-	}
-	batteryPercent = uint8(totalCharge / totalCapacity * 100)
-	batteryState = uint8(batteries[0].State.Raw)
-	return batteryPercent, batteryState, nil
-}

beszel/internal/agent/disk.go

@@ -1,190 +0,0 @@
-package agent
-
-import (
-	"beszel/internal/entities/system"
-	"log/slog"
-	"os"
-	"path/filepath"
-	"runtime"
-	"strings"
-	"time"
-
-	"github.com/shirou/gopsutil/v4/disk"
-)
-
-// Sets up the filesystems to monitor for disk usage and I/O.
-func (a *Agent) initializeDiskInfo() {
-	filesystem, _ := GetEnv("FILESYSTEM")
-	efPath := "/extra-filesystems"
-	hasRoot := false
-
-	partitions, err := disk.Partitions(false)
-	if err != nil {
-		slog.Error("Error getting disk partitions", "err", err)
-	}
-	slog.Debug("Disk", "partitions", partitions)
-
-	// ioContext := context.WithValue(a.sensorsContext,
-	// 	common.EnvKey, common.EnvMap{common.HostProcEnvKey: "/tmp/testproc"},
-	// )
-	// diskIoCounters, err := disk.IOCountersWithContext(ioContext)
-
-	diskIoCounters, err := disk.IOCounters()
-	if err != nil {
-		slog.Error("Error getting diskstats", "err", err)
-	}
-	slog.Debug("Disk I/O", "diskstats", diskIoCounters)
-
-	// Helper function to add a filesystem to fsStats if it doesn't exist
-	addFsStat := func(device, mountpoint string, root bool) {
-		var key string
-		if runtime.GOOS == "windows" {
-			key = device
-		} else {
-			key = filepath.Base(device)
-		}
-		var ioMatch bool
-		if _, exists := a.fsStats[key]; !exists {
-			if root {
-				slog.Info("Detected root device", "name", key)
-				// Check if root device is in /proc/diskstats, use fallback if not
-				if _, ioMatch = diskIoCounters[key]; !ioMatch {
-					key, ioMatch = findIoDevice(filesystem, diskIoCounters, a.fsStats)
-					if !ioMatch {
-						slog.Info("Using I/O fallback", "device", device, "mountpoint", mountpoint, "fallback", key)
-					}
-				}
-			} else {
-				// Check if non-root has diskstats and fall back to folder name if not
-				// Scenario: device is encrypted and named luks-2bcb02be-999d-4417-8d18-5c61e660fb6e - not in /proc/diskstats.
-				// However, the device can be specified by mounting folder from luks device at /extra-filesystems/sda1
-				if _, ioMatch = diskIoCounters[key]; !ioMatch {
-					efBase := filepath.Base(mountpoint)
-					if _, ioMatch = diskIoCounters[efBase]; ioMatch {
-						key = efBase
-					}
-				}
-			}
-			a.fsStats[key] = &system.FsStats{Root: root, Mountpoint: mountpoint}
-		}
-	}
-
-	// Use FILESYSTEM env var to find root filesystem
-	if filesystem != "" {
-		for _, p := range partitions {
-			if strings.HasSuffix(p.Device, filesystem) || p.Mountpoint == filesystem {
-				addFsStat(p.Device, p.Mountpoint, true)
-				hasRoot = true
-				break
-			}
-		}
-		if !hasRoot {
-			slog.Warn("Partition details not found", "filesystem", filesystem)
-		}
-	}
-
-	// Add EXTRA_FILESYSTEMS env var values to fsStats
-	if extraFilesystems, exists := GetEnv("EXTRA_FILESYSTEMS"); exists {
-		for _, fs := range strings.Split(extraFilesystems, ",") {
-			found := false
-			for _, p := range partitions {
-				if strings.HasSuffix(p.Device, fs) || p.Mountpoint == fs {
-					addFsStat(p.Device, p.Mountpoint, false)
-					found = true
-					break
-				}
-			}
-			// if not in partitions, test if we can get disk usage
-			if !found {
-				if _, err := disk.Usage(fs); err == nil {
-					addFsStat(filepath.Base(fs), fs, false)
-				} else {
-					slog.Error("Invalid filesystem", "name", fs, "err", err)
-				}
-			}
-		}
-	}
-
-	// Process partitions for various mount points
-	for _, p := range partitions {
-		// fmt.Println(p.Device, p.Mountpoint)
-		// Binary root fallback or docker root fallback
-		if !hasRoot && (p.Mountpoint == "/" || (p.Mountpoint == "/etc/hosts" && strings.HasPrefix(p.Device, "/dev"))) {
-			fs, match := findIoDevice(filepath.Base(p.Device), diskIoCounters, a.fsStats)
-			if match {
-				addFsStat(fs, p.Mountpoint, true)
-				hasRoot = true
-			}
-		}
-
-		// Check if device is in /extra-filesystems
-		if strings.HasPrefix(p.Mountpoint, efPath) {
-			addFsStat(p.Device, p.Mountpoint, false)
-		}
-	}
-
-	// Check all folders in /extra-filesystems and add them if not already present
-	if folders, err := os.ReadDir(efPath); err == nil {
-		existingMountpoints := make(map[string]bool)
-		for _, stats := range a.fsStats {
-			existingMountpoints[stats.Mountpoint] = true
-		}
-		for _, folder := range folders {
-			if folder.IsDir() {
-				mountpoint := filepath.Join(efPath, folder.Name())
-				slog.Debug("/extra-filesystems", "mountpoint", mountpoint)
-				if !existingMountpoints[mountpoint] {
-					addFsStat(folder.Name(), mountpoint, false)
-				}
-			}
-		}
-	}
-
-	// If no root filesystem set, use fallback
-	if !hasRoot {
-		rootDevice, _ := findIoDevice(filepath.Base(filesystem), diskIoCounters, a.fsStats)
-		slog.Info("Root disk", "mountpoint", "/", "io", rootDevice)
-		a.fsStats[rootDevice] = &system.FsStats{Root: true, Mountpoint: "/"}
-	}
-
-	a.initializeDiskIoStats(diskIoCounters)
-}
-
-// Returns matching device from /proc/diskstats,
-// or the device with the most reads if no match is found.
-// bool is true if a match was found.
-func findIoDevice(filesystem string, diskIoCounters map[string]disk.IOCountersStat, fsStats map[string]*system.FsStats) (string, bool) {
-	var maxReadBytes uint64
-	maxReadDevice := "/"
-	for _, d := range diskIoCounters {
-		if d.Name == filesystem || (d.Label != "" && d.Label == filesystem) {
-			return d.Name, true
-		}
-		if d.ReadBytes > maxReadBytes {
-			// don't use if device already exists in fsStats
-			if _, exists := fsStats[d.Name]; !exists {
-				maxReadBytes = d.ReadBytes
-				maxReadDevice = d.Name
-			}
-		}
-	}
-	return maxReadDevice, false
-}
-
-// Sets start values for disk I/O stats.
-func (a *Agent) initializeDiskIoStats(diskIoCounters map[string]disk.IOCountersStat) {
-	for device, stats := range a.fsStats {
-		// skip if not in diskIoCounters
-		d, exists := diskIoCounters[device]
-		if !exists {
-			slog.Warn("Device not found in diskstats", "name", device)
-			continue
-		}
-		// populate initial values
-		stats.Time = time.Now()
-		stats.TotalRead = d.ReadBytes
-		stats.TotalWrite = d.WriteBytes
-		// add to list of valid io device names
-		a.fsNames = append(a.fsNames, device)
-	}
-}

beszel/internal/agent/docker.go

@@ -1,369 +0,0 @@
-package agent
-
-import (
-	"beszel/internal/entities/container"
-	"bytes"
-	"context"
-	"encoding/json"
-	"fmt"
-	"log/slog"
-	"net"
-	"net/http"
-	"net/url"
-	"os"
-	"strings"
-	"sync"
-	"time"
-
-	"github.com/blang/semver"
-)
-
-type dockerManager struct {
-	client              *http.Client                // Client to query Docker API
-	wg                  sync.WaitGroup              // WaitGroup to wait for all goroutines to finish
-	sem                 chan struct{}               // Semaphore to limit concurrent container requests
-	containerStatsMutex sync.RWMutex                // Mutex to prevent concurrent access to containerStatsMap
-	apiContainerList    []*container.ApiInfo        // List of containers from Docker API (no pointer)
-	containerStatsMap   map[string]*container.Stats // Keeps track of container stats
-	validIds            map[string]struct{}         // Map of valid container ids, used to prune invalid containers from containerStatsMap
-	goodDockerVersion   bool                        // Whether docker version is at least 25.0.0 (one-shot works correctly)
-	isWindows           bool                        // Whether the Docker Engine API is running on Windows
-	buf                 *bytes.Buffer               // Buffer to store and read response bodies
-	decoder             *json.Decoder               // Reusable JSON decoder that reads from buf
-	apiStats            *container.ApiStats         // Reusable API stats object
-}
-
-// userAgentRoundTripper is a custom http.RoundTripper that adds a User-Agent header to all requests
-type userAgentRoundTripper struct {
-	rt        http.RoundTripper
-	userAgent string
-}
-
-// RoundTrip implements the http.RoundTripper interface
-func (u *userAgentRoundTripper) RoundTrip(req *http.Request) (*http.Response, error) {
-	req.Header.Set("User-Agent", u.userAgent)
-	return u.rt.RoundTrip(req)
-}
-
-// Add goroutine to the queue
-func (d *dockerManager) queue() {
-	d.wg.Add(1)
-	if d.goodDockerVersion {
-		d.sem <- struct{}{}
-	}
-}
-
-// Remove goroutine from the queue
-func (d *dockerManager) dequeue() {
-	d.wg.Done()
-	if d.goodDockerVersion {
-		<-d.sem
-	}
-}
-
-// Returns stats for all running containers
-func (dm *dockerManager) getDockerStats() ([]*container.Stats, error) {
-	resp, err := dm.client.Get("http://localhost/containers/json")
-	if err != nil {
-		return nil, err
-	}
-
-	dm.apiContainerList = dm.apiContainerList[:0]
-	if err := dm.decode(resp, &dm.apiContainerList); err != nil {
-		return nil, err
-	}
-
-	dm.isWindows = strings.Contains(resp.Header.Get("Server"), "windows")
-
-	containersLength := len(dm.apiContainerList)
-
-	// store valid ids to clean up old container ids from map
-	if dm.validIds == nil {
-		dm.validIds = make(map[string]struct{}, containersLength)
-	} else {
-		clear(dm.validIds)
-	}
-
-	var failedContainers []*container.ApiInfo
-
-	for i := range dm.apiContainerList {
-		ctr := dm.apiContainerList[i]
-		ctr.IdShort = ctr.Id[:12]
-		dm.validIds[ctr.IdShort] = struct{}{}
-		// check if container is less than 1 minute old (possible restart)
-		// note: can't use Created field because it's not updated on restart
-		if strings.Contains(ctr.Status, "second") {
-			// if so, remove old container data
-			dm.deleteContainerStatsSync(ctr.IdShort)
-		}
-		dm.queue()
-		go func() {
-			defer dm.dequeue()
-			err := dm.updateContainerStats(ctr)
-			// if error, delete from map and add to failed list to retry
-			if err != nil {
-				dm.containerStatsMutex.Lock()
-				delete(dm.containerStatsMap, ctr.IdShort)
-				failedContainers = append(failedContainers, ctr)
-				dm.containerStatsMutex.Unlock()
-			}
-		}()
-	}
-
-	dm.wg.Wait()
-
-	// retry failed containers separately so we can run them in parallel (docker 24 bug)
-	if len(failedContainers) > 0 {
-		slog.Debug("Retrying failed containers", "count", len(failedContainers))
-		for i := range failedContainers {
-			ctr := failedContainers[i]
-			dm.queue()
-			go func() {
-				defer dm.dequeue()
-				err = dm.updateContainerStats(ctr)
-				if err != nil {
-					slog.Error("Error getting container stats", "err", err)
-				}
-			}()
-		}
-		dm.wg.Wait()
-	}
-
-	// populate final stats and remove old / invalid container stats
-	stats := make([]*container.Stats, 0, containersLength)
-	for id, v := range dm.containerStatsMap {
-		if _, exists := dm.validIds[id]; !exists {
-			delete(dm.containerStatsMap, id)
-		} else {
-			stats = append(stats, v)
-		}
-	}
-
-	return stats, nil
-}
-
-// Updates stats for individual container
-func (dm *dockerManager) updateContainerStats(ctr *container.ApiInfo) error {
-	name := ctr.Names[0][1:]
-
-	resp, err := dm.client.Get("http://localhost/containers/" + ctr.IdShort + "/stats?stream=0&one-shot=1")
-	if err != nil {
-		return err
-	}
-	defer resp.Body.Close()
-
-	dm.containerStatsMutex.Lock()
-	defer dm.containerStatsMutex.Unlock()
-
-	// add empty values if they doesn't exist in map
-	stats, initialized := dm.containerStatsMap[ctr.IdShort]
-	if !initialized {
-		stats = &container.Stats{Name: name}
-		dm.containerStatsMap[ctr.IdShort] = stats
-	}
-
-	// reset current stats
-	stats.Cpu = 0
-	stats.Mem = 0
-	stats.NetworkSent = 0
-	stats.NetworkRecv = 0
-
-	// docker host container stats response
-	// res := dm.getApiStats()
-	// defer dm.putApiStats(res)
-	//
-
-	res := dm.apiStats
-	res.Networks = nil
-	if err := dm.decode(resp, res); err != nil {
-		return err
-	}
-
-	// calculate cpu and memory stats
-	var usedMemory uint64
-	var cpuPct float64
-
-	// store current cpu stats
-	prevCpuContainer, prevCpuSystem := stats.CpuContainer, stats.CpuSystem
-	stats.CpuContainer = res.CPUStats.CPUUsage.TotalUsage
-	stats.CpuSystem = res.CPUStats.SystemUsage
-
-	if dm.isWindows {
-		usedMemory = res.MemoryStats.PrivateWorkingSet
-		cpuPct = res.CalculateCpuPercentWindows(prevCpuContainer, stats.PrevReadTime)
-	} else {
-		// check if container has valid data, otherwise may be in restart loop (#103)
-		if res.MemoryStats.Usage == 0 {
-			return fmt.Errorf("%s - no memory stats - see https://github.com/henrygd/beszel/issues/144", name)
-		}
-		memCache := res.MemoryStats.Stats.InactiveFile
-		if memCache == 0 {
-			memCache = res.MemoryStats.Stats.Cache
-		}
-		usedMemory = res.MemoryStats.Usage - memCache
-
-		cpuPct = res.CalculateCpuPercentLinux(prevCpuContainer, prevCpuSystem)
-	}
-
-	if cpuPct > 100 {
-		return fmt.Errorf("%s cpu pct greater than 100: %+v", name, cpuPct)
-	}
-
-	// network
-	var total_sent, total_recv uint64
-	for _, v := range res.Networks {
-		total_sent += v.TxBytes
-		total_recv += v.RxBytes
-	}
-	var sent_delta, recv_delta uint64
-	millisecondsElapsed := uint64(time.Since(stats.PrevReadTime).Milliseconds())
-	if initialized && millisecondsElapsed > 0 {
-		// get bytes per second
-		sent_delta = (total_sent - stats.PrevNet.Sent) * 1000 / millisecondsElapsed
-		recv_delta = (total_recv - stats.PrevNet.Recv) * 1000 / millisecondsElapsed
-		// check for unrealistic network values (> 5GB/s)
-		if sent_delta > 5e9 || recv_delta > 5e9 {
-			slog.Warn("Bad network delta", "container", name)
-			sent_delta, recv_delta = 0, 0
-		}
-	}
-	stats.PrevNet.Sent, stats.PrevNet.Recv = total_sent, total_recv
-
-	stats.Cpu = twoDecimals(cpuPct)
-	stats.Mem = bytesToMegabytes(float64(usedMemory))
-	stats.NetworkSent = bytesToMegabytes(float64(sent_delta))
-	stats.NetworkRecv = bytesToMegabytes(float64(recv_delta))
-	stats.PrevReadTime = res.Read
-
-	return nil
-}
-
-// Delete container stats from map using mutex
-func (dm *dockerManager) deleteContainerStatsSync(id string) {
-	dm.containerStatsMutex.Lock()
-	defer dm.containerStatsMutex.Unlock()
-	delete(dm.containerStatsMap, id)
-}
-
-// Creates a new http client for Docker or Podman API
-func newDockerManager(a *Agent) *dockerManager {
-	dockerHost, exists := GetEnv("DOCKER_HOST")
-	if exists {
-		// return nil if set to empty string
-		if dockerHost == "" {
-			return nil
-		}
-	} else {
-		dockerHost = getDockerHost()
-	}
-
-	parsedURL, err := url.Parse(dockerHost)
-	if err != nil {
-		os.Exit(1)
-	}
-
-	transport := &http.Transport{
-		DisableCompression: true,
-		MaxConnsPerHost:    0,
-	}
-
-	switch parsedURL.Scheme {
-	case "unix":
-		transport.DialContext = func(ctx context.Context, proto, addr string) (net.Conn, error) {
-			return (&net.Dialer{}).DialContext(ctx, "unix", parsedURL.Path)
-		}
-	case "tcp", "http", "https":
-		transport.DialContext = func(ctx context.Context, proto, addr string) (net.Conn, error) {
-			return (&net.Dialer{}).DialContext(ctx, "tcp", parsedURL.Host)
-		}
-	default:
-		slog.Error("Invalid DOCKER_HOST", "scheme", parsedURL.Scheme)
-		os.Exit(1)
-	}
-
-	// configurable timeout
-	timeout := time.Millisecond * 2100
-	if t, set := GetEnv("DOCKER_TIMEOUT"); set {
-		timeout, err = time.ParseDuration(t)
-		if err != nil {
-			slog.Error(err.Error())
-			os.Exit(1)
-		}
-		slog.Info("DOCKER_TIMEOUT", "timeout", timeout)
-	}
-
-	// Custom user-agent to avoid docker bug: https://github.com/docker/for-mac/issues/7575
-	userAgentTransport := &userAgentRoundTripper{
-		rt:        transport,
-		userAgent: "Docker-Client/",
-	}
-
-	manager := &dockerManager{
-		client: &http.Client{
-			Timeout:   timeout,
-			Transport: userAgentTransport,
-		},
-		containerStatsMap: make(map[string]*container.Stats),
-		sem:               make(chan struct{}, 5),
-		apiContainerList:  []*container.ApiInfo{},
-		apiStats:          &container.ApiStats{},
-	}
-
-	// If using podman, return client
-	if strings.Contains(dockerHost, "podman") {
-		a.systemInfo.Podman = true
-		manager.goodDockerVersion = true
-		return manager
-	}
-
-	// Check docker version
-	// (versions before 25.0.0 have a bug with one-shot which requires all requests to be made in one batch)
-	var versionInfo struct {
-		Version string `json:"Version"`
-	}
-	resp, err := manager.client.Get("http://localhost/version")
-	if err != nil {
-		return manager
-	}
-
-	if err := manager.decode(resp, &versionInfo); err != nil {
-		return manager
-	}
-
-	// if version > 24, one-shot works correctly and we can limit concurrent operations
-	if dockerVersion, err := semver.Parse(versionInfo.Version); err == nil && dockerVersion.Major > 24 {
-		manager.goodDockerVersion = true
-	} else {
-		slog.Info(fmt.Sprintf("Docker %s is outdated. Upgrade if possible. See https://github.com/henrygd/beszel/issues/58", versionInfo.Version))
-	}
-
-	return manager
-}
-
-// Decodes Docker API JSON response using a reusable buffer and decoder. Not thread safe.
-func (dm *dockerManager) decode(resp *http.Response, d any) error {
-	if dm.buf == nil {
-		// initialize buffer with 256kb starting size
-		dm.buf = bytes.NewBuffer(make([]byte, 0, 1024*256))
-		dm.decoder = json.NewDecoder(dm.buf)
-	}
-	defer resp.Body.Close()
-	defer dm.buf.Reset()
-	_, err := dm.buf.ReadFrom(resp.Body)
-	if err != nil {
-		return err
-	}
-	return dm.decoder.Decode(d)
-}
-
-// Test docker / podman sockets and return if one exists
-func getDockerHost() string {
-	scheme := "unix://"
-	socks := []string{"/var/run/docker.sock", fmt.Sprintf("/run/user/%v/podman/podman.sock", os.Getuid())}
-	for _, sock := range socks {
-		if _, err := os.Stat(sock); err == nil {
-			return scheme + sock
-		}
-	}
-	return scheme + socks[0]
-}

beszel/internal/agent/gpu_test.go

@@ -1,793 +0,0 @@
-//go:build testing
-// +build testing
-
-package agent
-
-import (
-	"beszel/internal/entities/system"
-	"os"
-	"path/filepath"
-	"testing"
-	"time"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-func TestParseNvidiaData(t *testing.T) {
-	tests := []struct {
-		name      string
-		input     string
-		wantData  map[string]system.GPUData
-		wantValid bool
-	}{
-		{
-			name:  "valid multi-gpu data",
-			input: "0, NVIDIA GeForce RTX 3050 Ti Laptop GPU, 48, 12, 4096, 26.3, 12.73\n1, NVIDIA A100-PCIE-40GB, 38, 74, 40960, [N/A], 36.79",
-			wantData: map[string]system.GPUData{
-				"0": {
-					Name:        "GeForce RTX 3050 Ti",
-					Temperature: 48.0,
-					MemoryUsed:  12.0 / 1.024,
-					MemoryTotal: 4096.0 / 1.024,
-					Usage:       26.3,
-					Power:       12.73,
-					Count:       1,
-				},
-				"1": {
-					Name:        "A100-PCIE-40GB",
-					Temperature: 38.0,
-					MemoryUsed:  74.0 / 1.024,
-					MemoryTotal: 40960.0 / 1.024,
-					Usage:       0.0,
-					Power:       36.79,
-					Count:       1,
-				},
-			},
-			wantValid: true,
-		},
-		{
-			name: "more valid multi-gpu data",
-			input: `0, NVIDIA A10, 45, 19676, 23028, 0, 58.98
-1, NVIDIA A10, 45, 19638, 23028, 0, 62.35
-2, NVIDIA A10, 44, 21700, 23028, 0, 59.57
-3, NVIDIA A10, 45, 18222, 23028, 0, 61.76`,
-			wantData: map[string]system.GPUData{
-				"0": {
-					Name:        "A10",
-					Temperature: 45.0,
-					MemoryUsed:  19676.0 / 1.024,
-					MemoryTotal: 23028.0 / 1.024,
-					Usage:       0.0,
-					Power:       58.98,
-					Count:       1,
-				},
-				"1": {
-					Name:        "A10",
-					Temperature: 45.0,
-					MemoryUsed:  19638.0 / 1.024,
-					MemoryTotal: 23028.0 / 1.024,
-					Usage:       0.0,
-					Power:       62.35,
-					Count:       1,
-				},
-				"2": {
-					Name:        "A10",
-					Temperature: 44.0,
-					MemoryUsed:  21700.0 / 1.024,
-					MemoryTotal: 23028.0 / 1.024,
-					Usage:       0.0,
-					Power:       59.57,
-					Count:       1,
-				},
-				"3": {
-					Name:        "A10",
-					Temperature: 45.0,
-					MemoryUsed:  18222.0 / 1.024,
-					MemoryTotal: 23028.0 / 1.024,
-					Usage:       0.0,
-					Power:       61.76,
-					Count:       1,
-				},
-			},
-			wantValid: true,
-		},
-		{
-			name:      "empty input",
-			input:     "",
-			wantData:  map[string]system.GPUData{},
-			wantValid: false,
-		},
-		{
-			name:      "malformed data",
-			input:     "bad, data, here",
-			wantData:  map[string]system.GPUData{},
-			wantValid: false,
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			gm := &GPUManager{
-				GpuDataMap: make(map[string]*system.GPUData),
-			}
-			valid := gm.parseNvidiaData([]byte(tt.input))
-			assert.Equal(t, tt.wantValid, valid)
-
-			if tt.wantValid {
-				for id, want := range tt.wantData {
-					got := gm.GpuDataMap[id]
-					require.NotNil(t, got)
-					assert.Equal(t, want.Name, got.Name)
-					assert.InDelta(t, want.Temperature, got.Temperature, 0.01)
-					assert.InDelta(t, want.MemoryUsed, got.MemoryUsed, 0.01)
-					assert.InDelta(t, want.MemoryTotal, got.MemoryTotal, 0.01)
-					assert.InDelta(t, want.Usage, got.Usage, 0.01)
-					assert.InDelta(t, want.Power, got.Power, 0.01)
-					assert.Equal(t, want.Count, got.Count)
-				}
-			}
-		})
-	}
-}
-
-func TestParseAmdData(t *testing.T) {
-	tests := []struct {
-		name      string
-		input     string
-		wantData  map[string]system.GPUData
-		wantValid bool
-	}{
-		{
-			name: "valid single gpu data",
-			input: `{
-				"card0": {
-					"GUID": "34756",
-					"Temperature (Sensor edge) (C)": "47.0",
-					"Current Socket Graphics Package Power (W)": "9.215",
-					"GPU use (%)": "0",
-					"VRAM Total Memory (B)": "536870912",
-					"VRAM Total Used Memory (B)": "482263040",
-					"Card Series": "Rembrandt [Radeon 680M]"
-				}
-			}`,
-			wantData: map[string]system.GPUData{
-				"34756": {
-					Name:        "Rembrandt [Radeon 680M]",
-					Temperature: 47.0,
-					MemoryUsed:  482263040.0 / (1024 * 1024),
-					MemoryTotal: 536870912.0 / (1024 * 1024),
-					Usage:       0.0,
-					Power:       9.215,
-					Count:       1,
-				},
-			},
-			wantValid: true,
-		},
-		{
-			name: "valid multi gpu data",
-			input: `{
-				"card0": {
-					"GUID": "34756",
-					"Temperature (Sensor edge) (C)": "47.0",
-					"Current Socket Graphics Package Power (W)": "9.215",
-					"GPU use (%)": "0",
-					"VRAM Total Memory (B)": "536870912",
-					"VRAM Total Used Memory (B)": "482263040",
-					"Card Series": "Rembrandt [Radeon 680M]"
-				},
-				"card1": {
-					"GUID": "38294",
-					"Temperature (Sensor edge) (C)": "49.0",
-					"Temperature (Sensor junction) (C)": "49.0",
-					"Temperature (Sensor memory) (C)": "62.0",
-					"Average Graphics Package Power (W)": "19.0",
-					"GPU use (%)": "20.3",
-					"VRAM Total Memory (B)": "25753026560",
-					"VRAM Total Used Memory (B)": "794341376",
-					"Card Series": "Navi 31 [Radeon RX 7900 XT]"
-				}
-			}`,
-			wantData: map[string]system.GPUData{
-				"34756": {
-					Name:        "Rembrandt [Radeon 680M]",
-					Temperature: 47.0,
-					MemoryUsed:  482263040.0 / (1024 * 1024),
-					MemoryTotal: 536870912.0 / (1024 * 1024),
-					Usage:       0.0,
-					Power:       9.215,
-					Count:       1,
-				},
-				"38294": {
-					Name:        "Navi 31 [Radeon RX 7900 XT]",
-					Temperature: 49.0,
-					MemoryUsed:  794341376.0 / (1024 * 1024),
-					MemoryTotal: 25753026560.0 / (1024 * 1024),
-					Usage:       20.3,
-					Power:       19.0,
-					Count:       1,
-				},
-			},
-			wantValid: true,
-		},
-		{
-			name:  "invalid json",
-			input: "{bad json",
-		},
-		{
-			name:      "invalid json",
-			input:     "{bad json",
-			wantData:  map[string]system.GPUData{},
-			wantValid: false,
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			gm := &GPUManager{
-				GpuDataMap: make(map[string]*system.GPUData),
-			}
-			valid := gm.parseAmdData([]byte(tt.input))
-			assert.Equal(t, tt.wantValid, valid)
-
-			if tt.wantValid {
-				for id, want := range tt.wantData {
-					got := gm.GpuDataMap[id]
-					require.NotNil(t, got)
-					assert.Equal(t, want.Name, got.Name)
-					assert.InDelta(t, want.Temperature, got.Temperature, 0.01)
-					assert.InDelta(t, want.MemoryUsed, got.MemoryUsed, 0.01)
-					assert.InDelta(t, want.MemoryTotal, got.MemoryTotal, 0.01)
-					assert.InDelta(t, want.Usage, got.Usage, 0.01)
-					assert.InDelta(t, want.Power, got.Power, 0.01)
-					assert.Equal(t, want.Count, got.Count)
-				}
-			}
-		})
-	}
-}
-
-func TestParseJetsonData(t *testing.T) {
-	tests := []struct {
-		name        string
-		input       string
-		wantMetrics *system.GPUData
-	}{
-		{
-			name:  "valid data",
-			input: "11-14-2024 22:54:33 RAM 4300/30698MB GR3D_FREQ 45% tj@52.468C VDD_GPU_SOC 2171mW",
-			wantMetrics: &system.GPUData{
-				Name:        "GPU",
-				MemoryUsed:  4300.0,
-				MemoryTotal: 30698.0,
-				Usage:       45.0,
-				Temperature: 52.468,
-				Power:       2.171,
-				Count:       1,
-			},
-		},
-		{
-			name:  "more valid data",
-			input: "11-15-2024 08:38:09 RAM 6185/7620MB (lfb 8x2MB) SWAP 851/3810MB (cached 1MB) CPU [15%@729,11%@729,14%@729,13%@729,11%@729,8%@729] EMC_FREQ 43%@2133 GR3D_FREQ 63%@[621] NVDEC off NVJPG off NVJPG1 off VIC off OFA off APE 200 cpu@53.968C soc2@52.437C soc0@50.75C gpu@53.343C tj@53.968C soc1@51.656C VDD_IN 12479mW/12479mW VDD_CPU_GPU_CV 4667mW/4667mW VDD_SOC 2817mW/2817mW",
-			wantMetrics: &system.GPUData{
-				Name:        "GPU",
-				MemoryUsed:  6185.0,
-				MemoryTotal: 7620.0,
-				Usage:       63.0,
-				Temperature: 53.968,
-				Power:       4.667,
-				Count:       1,
-			},
-		},
-		{
-			name:  "orin nano",
-			input: "06-18-2025 11:25:24 RAM 3452/7620MB (lfb 25x4MB) SWAP 1518/16384MB (cached 174MB) CPU [1%@1420,2%@1420,0%@1420,2%@1420,2%@729,1%@729] GR3D_FREQ 0% cpu@50.031C soc2@49.031C soc0@50C gpu@49.031C tj@50.25C soc1@50.25C VDD_IN 4824mW/4824mW VDD_CPU_GPU_CV 518mW/518mW VDD_SOC 1475mW/1475mW",
-			wantMetrics: &system.GPUData{
-				Name:        "GPU",
-				MemoryUsed:  3452.0,
-				MemoryTotal: 7620.0,
-				Usage:       0.0,
-				Temperature: 50.25,
-				Power:       0.518,
-				Count:       1,
-			},
-		},
-		{
-			name:  "missing temperature",
-			input: "11-14-2024 22:54:33 RAM 4300/30698MB GR3D_FREQ 45% VDD_GPU_SOC 2171mW",
-			wantMetrics: &system.GPUData{
-				Name:        "GPU",
-				MemoryUsed:  4300.0,
-				MemoryTotal: 30698.0,
-				Usage:       45.0,
-				Power:       2.171,
-				Count:       1,
-			},
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			gm := &GPUManager{
-				GpuDataMap: make(map[string]*system.GPUData),
-			}
-			parser := gm.getJetsonParser()
-			valid := parser([]byte(tt.input))
-			assert.Equal(t, true, valid)
-
-			got := gm.GpuDataMap["0"]
-			require.NotNil(t, got)
-			assert.Equal(t, tt.wantMetrics.Name, got.Name)
-			assert.InDelta(t, tt.wantMetrics.MemoryUsed, got.MemoryUsed, 0.01)
-			assert.InDelta(t, tt.wantMetrics.MemoryTotal, got.MemoryTotal, 0.01)
-			assert.InDelta(t, tt.wantMetrics.Usage, got.Usage, 0.01)
-			if tt.wantMetrics.Temperature > 0 {
-				assert.InDelta(t, tt.wantMetrics.Temperature, got.Temperature, 0.01)
-			}
-			assert.InDelta(t, tt.wantMetrics.Power, got.Power, 0.01)
-			assert.Equal(t, tt.wantMetrics.Count, got.Count)
-		})
-	}
-}
-
-func TestGetCurrentData(t *testing.T) {
-	t.Run("calculates averages and resets accumulators", func(t *testing.T) {
-		gm := &GPUManager{
-			GpuDataMap: map[string]*system.GPUData{
-				"0": {
-					Name:        "GPU1",
-					Temperature: 50,
-					MemoryUsed:  2048,
-					MemoryTotal: 4096,
-					Usage:       100, // 100 over 2 counts = 50 avg
-					Power:       200, // 200 over 2 counts = 100 avg
-					Count:       2,
-				},
-				"1": {
-					Name:        "GPU1",
-					Temperature: 60,
-					MemoryUsed:  3072,
-					MemoryTotal: 8192,
-					Usage:       30,
-					Power:       60,
-					Count:       1,
-				},
-				"2": {
-					Name:        "GPU 2",
-					Temperature: 70,
-					MemoryUsed:  4096,
-					MemoryTotal: 8192,
-					Usage:       200,
-					Power:       400,
-					Count:       1,
-				},
-			},
-		}
-
-		result := gm.GetCurrentData()
-
-		// Verify name disambiguation
-		assert.Equal(t, "GPU1 0", result["0"].Name)
-		assert.Equal(t, "GPU1 1", result["1"].Name)
-		assert.Equal(t, "GPU 2", result["2"].Name)
-
-		// Check averaged values in the result
-		assert.InDelta(t, 50.0, result["0"].Usage, 0.01)
-		assert.InDelta(t, 100.0, result["0"].Power, 0.01)
-		assert.InDelta(t, 30.0, result["1"].Usage, 0.01)
-		assert.InDelta(t, 60.0, result["1"].Power, 0.01)
-
-		// Verify that accumulators in the original map are reset
-		assert.Equal(t, float64(0), gm.GpuDataMap["0"].Count, "GPU 0 Count should be reset")
-		assert.Equal(t, float64(0), gm.GpuDataMap["0"].Usage, "GPU 0 Usage should be reset")
-		assert.Equal(t, float64(0), gm.GpuDataMap["0"].Power, "GPU 0 Power should be reset")
-		assert.Equal(t, float64(0), gm.GpuDataMap["1"].Count, "GPU 1 Count should be reset")
-		assert.Equal(t, float64(0), gm.GpuDataMap["1"].Usage, "GPU 1 Usage should be reset")
-		assert.Equal(t, float64(0), gm.GpuDataMap["1"].Power, "GPU 1 Power should be reset")
-	})
-
-	t.Run("handles zero count without panicking", func(t *testing.T) {
-		gm := &GPUManager{
-			GpuDataMap: map[string]*system.GPUData{
-				"0": {
-					Name:  "TestGPU",
-					Count: 0,
-					Usage: 0,
-					Power: 0,
-				},
-			},
-		}
-
-		var result map[string]system.GPUData
-		assert.NotPanics(t, func() {
-			result = gm.GetCurrentData()
-		})
-
-		// Check that usage and power are 0
-		assert.Equal(t, 0.0, result["0"].Usage)
-		assert.Equal(t, 0.0, result["0"].Power)
-
-		// Verify reset count
-		assert.Equal(t, float64(0), gm.GpuDataMap["0"].Count)
-	})
-}
-
-func TestDetectGPUs(t *testing.T) {
-	// Save original PATH
-	origPath := os.Getenv("PATH")
-	defer os.Setenv("PATH", origPath)
-
-	// Set up temp dir with the commands
-	tempDir := t.TempDir()
-	os.Setenv("PATH", tempDir)
-
-	tests := []struct {
-		name           string
-		setupCommands  func() error
-		wantNvidiaSmi  bool
-		wantRocmSmi    bool
-		wantTegrastats bool
-		wantErr        bool
-	}{
-		{
-			name: "nvidia-smi not available",
-			setupCommands: func() error {
-				return nil
-			},
-			wantNvidiaSmi:  false,
-			wantRocmSmi:    false,
-			wantTegrastats: false,
-			wantErr:        true,
-		},
-		{
-			name: "nvidia-smi available",
-			setupCommands: func() error {
-				path := filepath.Join(tempDir, "nvidia-smi")
-				script := `#!/bin/sh
-echo "test"`
-				if err := os.WriteFile(path, []byte(script), 0755); err != nil {
-					return err
-				}
-				return nil
-			},
-			wantNvidiaSmi:  true,
-			wantTegrastats: false,
-			wantRocmSmi:    false,
-			wantErr:        false,
-		},
-		{
-			name: "rocm-smi available",
-			setupCommands: func() error {
-				path := filepath.Join(tempDir, "rocm-smi")
-				script := `#!/bin/sh
-echo "test"`
-				if err := os.WriteFile(path, []byte(script), 0755); err != nil {
-					return err
-				}
-				return nil
-			},
-			wantNvidiaSmi:  true,
-			wantRocmSmi:    true,
-			wantTegrastats: false,
-			wantErr:        false,
-		},
-		{
-			name: "tegrastats available",
-			setupCommands: func() error {
-				path := filepath.Join(tempDir, "tegrastats")
-				script := `#!/bin/sh
-echo "test"`
-				if err := os.WriteFile(path, []byte(script), 0755); err != nil {
-					return err
-				}
-				return nil
-			},
-			wantNvidiaSmi:  false,
-			wantRocmSmi:    true,
-			wantTegrastats: true,
-			wantErr:        false,
-		},
-		{
-			name: "no gpu tools available",
-			setupCommands: func() error {
-				os.Setenv("PATH", "")
-				return nil
-			},
-			wantErr: true,
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			if err := tt.setupCommands(); err != nil {
-				t.Fatal(err)
-			}
-
-			gm := &GPUManager{}
-			err := gm.detectGPUs()
-
-			t.Logf("nvidiaSmi: %v, rocmSmi: %v, tegrastats: %v", gm.nvidiaSmi, gm.rocmSmi, gm.tegrastats)
-
-			if tt.wantErr {
-				assert.Error(t, err)
-				return
-			}
-
-			assert.NoError(t, err)
-			assert.Equal(t, tt.wantNvidiaSmi, gm.nvidiaSmi)
-			assert.Equal(t, tt.wantRocmSmi, gm.rocmSmi)
-			assert.Equal(t, tt.wantTegrastats, gm.tegrastats)
-		})
-	}
-}
-
-func TestStartCollector(t *testing.T) {
-	// Save original PATH
-	origPath := os.Getenv("PATH")
-	defer os.Setenv("PATH", origPath)
-
-	// Set up temp dir with the commands
-	dir := t.TempDir()
-	os.Setenv("PATH", dir)
-
-	tests := []struct {
-		name     string
-		command  string
-		setup    func(t *testing.T) error
-		validate func(t *testing.T, gm *GPUManager)
-		gm       *GPUManager
-	}{
-		{
-			name:    "nvidia-smi collector",
-			command: "nvidia-smi",
-			setup: func(t *testing.T) error {
-				path := filepath.Join(dir, "nvidia-smi")
-				script := `#!/bin/sh
-echo "0, NVIDIA Test GPU, 50, 1024, 4096, 25, 100"`
-				if err := os.WriteFile(path, []byte(script), 0755); err != nil {
-					return err
-				}
-				return nil
-			},
-			validate: func(t *testing.T, gm *GPUManager) {
-				gpu, exists := gm.GpuDataMap["0"]
-				assert.True(t, exists)
-				if exists {
-					assert.Equal(t, "Test GPU", gpu.Name)
-					assert.Equal(t, 50.0, gpu.Temperature)
-
-				}
-			},
-		},
-		{
-			name:    "rocm-smi collector",
-			command: "rocm-smi",
-			setup: func(t *testing.T) error {
-				path := filepath.Join(dir, "rocm-smi")
-				script := `#!/bin/sh
-echo '{"card0": {"Temperature (Sensor edge) (C)": "49.0", "Current Socket Graphics Package Power (W)": "28.159", "GPU use (%)": "0", "VRAM Total Memory (B)": "536870912", "VRAM Total Used Memory (B)": "445550592", "Card Series": "Rembrandt [Radeon 680M]", "Card Model": "0x1681", "Card Vendor": "Advanced Micro Devices, Inc. [AMD/ATI]", "Card SKU": "REMBRANDT", "Subsystem ID": "0x8a22", "Device Rev": "0xc8", "Node ID": "1", "GUID": "34756", "GFX Version": "gfx1035"}}'`
-				if err := os.WriteFile(path, []byte(script), 0755); err != nil {
-					return err
-				}
-				return nil
-			},
-			validate: func(t *testing.T, gm *GPUManager) {
-				gpu, exists := gm.GpuDataMap["34756"]
-				assert.True(t, exists)
-				if exists {
-					assert.Equal(t, "Rembrandt [Radeon 680M]", gpu.Name)
-					assert.InDelta(t, 49.0, gpu.Temperature, 0.01)
-					assert.InDelta(t, 28.159, gpu.Power, 0.01)
-				}
-			},
-		},
-		{
-			name:    "tegrastats collector",
-			command: "tegrastats",
-			setup: func(t *testing.T) error {
-				path := filepath.Join(dir, "tegrastats")
-				script := `#!/bin/sh
-echo "11-14-2024 22:54:33 RAM 1024/4096MB GR3D_FREQ 80% tj@70C VDD_GPU_SOC 1000mW"`
-				if err := os.WriteFile(path, []byte(script), 0755); err != nil {
-					return err
-				}
-				return nil
-			},
-			validate: func(t *testing.T, gm *GPUManager) {
-				gpu, exists := gm.GpuDataMap["0"]
-				assert.True(t, exists)
-				if exists {
-					assert.InDelta(t, 70.0, gpu.Temperature, 0.1)
-				}
-			},
-			gm: &GPUManager{
-				GpuDataMap: map[string]*system.GPUData{
-					"0": {},
-				},
-			},
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			if err := tt.setup(t); err != nil {
-				t.Fatal(err)
-			}
-			if tt.gm == nil {
-				tt.gm = &GPUManager{
-					GpuDataMap: make(map[string]*system.GPUData),
-				}
-			}
-			tt.gm.startCollector(tt.command)
-			time.Sleep(50 * time.Millisecond) // Give collector time to run
-			tt.validate(t, tt.gm)
-		})
-	}
-}
-
-// TestAccumulationTableDriven tests the accumulation behavior for all three GPU types
-func TestAccumulation(t *testing.T) {
-	type expectedGPUValues struct {
-		temperature float64
-		memoryUsed  float64
-		memoryTotal float64
-		usage       float64
-		power       float64
-		count       float64
-		avgUsage    float64
-		avgPower    float64
-	}
-
-	tests := []struct {
-		name           string
-		initialGPUData map[string]*system.GPUData
-		dataSamples    [][]byte
-		parser         func(*GPUManager) func([]byte) bool
-		expectedValues map[string]expectedGPUValues
-	}{
-		{
-			name: "Jetson GPU accumulation",
-			initialGPUData: map[string]*system.GPUData{
-				"0": {
-					Name:        "Jetson",
-					Temperature: 0,
-					Usage:       0,
-					Power:       0,
-					Count:       0,
-				},
-			},
-			dataSamples: [][]byte{
-				[]byte("11-14-2024 22:54:33 RAM 1024/4096MB GR3D_FREQ 30% tj@50.5C VDD_GPU_SOC 1000mW"),
-				[]byte("11-14-2024 22:54:33 RAM 1024/4096MB GR3D_FREQ 40% tj@60.5C VDD_GPU_SOC 1200mW"),
-				[]byte("11-14-2024 22:54:33 RAM 1024/4096MB GR3D_FREQ 50% tj@70.5C VDD_GPU_SOC 1400mW"),
-			},
-			parser: func(gm *GPUManager) func([]byte) bool {
-				return gm.getJetsonParser()
-			},
-			expectedValues: map[string]expectedGPUValues{
-				"0": {
-					temperature: 70.5,  // Last value
-					memoryUsed:  1024,  // Last value
-					memoryTotal: 4096,  // Last value
-					usage:       120.0, // Accumulated: 30 + 40 + 50
-					power:       3.6,   // Accumulated: 1.0 + 1.2 + 1.4
-					count:       3,
-					avgUsage:    40.0, // 120 / 3
-					avgPower:    1.2,  // 3.6 / 3
-				},
-			},
-		},
-		{
-			name:           "NVIDIA GPU accumulation",
-			initialGPUData: map[string]*system.GPUData{
-				// NVIDIA parser will create the GPU data entries
-			},
-			dataSamples: [][]byte{
-				[]byte("0, NVIDIA GeForce RTX 3080, 50, 5000, 10000, 30, 200"),
-				[]byte("0, NVIDIA GeForce RTX 3080, 60, 6000, 10000, 40, 250"),
-				[]byte("0, NVIDIA GeForce RTX 3080, 70, 7000, 10000, 50, 300"),
-			},
-			parser: func(gm *GPUManager) func([]byte) bool {
-				return gm.parseNvidiaData
-			},
-			expectedValues: map[string]expectedGPUValues{
-				"0": {
-					temperature: 70.0,            // Last value
-					memoryUsed:  7000.0 / 1.024,  // Last value
-					memoryTotal: 10000.0 / 1.024, // Last value
-					usage:       120.0,           // Accumulated: 30 + 40 + 50
-					power:       750.0,           // Accumulated: 200 + 250 + 300
-					count:       3,
-					avgUsage:    40.0,  // 120 / 3
-					avgPower:    250.0, // 750 / 3
-				},
-			},
-		},
-		{
-			name:           "AMD GPU accumulation",
-			initialGPUData: map[string]*system.GPUData{
-				// AMD parser will create the GPU data entries
-			},
-			dataSamples: [][]byte{
-				[]byte(`{"card0": {"GUID": "34756", "Temperature (Sensor edge) (C)": "50.0", "Current Socket Graphics Package Power (W)": "100.0", "GPU use (%)": "30", "VRAM Total Memory (B)": "10737418240", "VRAM Total Used Memory (B)": "1073741824", "Card Series": "Radeon RX 6800"}}`),
-				[]byte(`{"card0": {"GUID": "34756", "Temperature (Sensor edge) (C)": "60.0", "Current Socket Graphics Package Power (W)": "150.0", "GPU use (%)": "40", "VRAM Total Memory (B)": "10737418240", "VRAM Total Used Memory (B)": "2147483648", "Card Series": "Radeon RX 6800"}}`),
-				[]byte(`{"card0": {"GUID": "34756", "Temperature (Sensor edge) (C)": "70.0", "Current Socket Graphics Package Power (W)": "200.0", "GPU use (%)": "50", "VRAM Total Memory (B)": "10737418240", "VRAM Total Used Memory (B)": "3221225472", "Card Series": "Radeon RX 6800"}}`),
-			},
-			parser: func(gm *GPUManager) func([]byte) bool {
-				return gm.parseAmdData
-			},
-			expectedValues: map[string]expectedGPUValues{
-				"34756": {
-					temperature: 70.0,                          // Last value
-					memoryUsed:  3221225472.0 / (1024 * 1024),  // Last value
-					memoryTotal: 10737418240.0 / (1024 * 1024), // Last value
-					usage:       120.0,                         // Accumulated: 30 + 40 + 50
-					power:       450.0,                         // Accumulated: 100 + 150 + 200
-					count:       3,
-					avgUsage:    40.0,  // 120 / 3
-					avgPower:    150.0, // 450 / 3
-				},
-			},
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			// Create a new GPUManager for each test
-			gm := &GPUManager{
-				GpuDataMap: tt.initialGPUData,
-			}
-
-			// Get the parser function
-			parser := tt.parser(gm)
-
-			// Process each data sample
-			for i, sample := range tt.dataSamples {
-				valid := parser(sample)
-				assert.True(t, valid, "Sample %d should be valid", i)
-			}
-
-			// Check accumulated values
-			for id, expected := range tt.expectedValues {
-				gpu, exists := gm.GpuDataMap[id]
-				assert.True(t, exists, "GPU with ID %s should exist", id)
-				if !exists {
-					continue
-				}
-
-				assert.InDelta(t, expected.temperature, gpu.Temperature, 0.01, "Temperature should match")
-				assert.InDelta(t, expected.memoryUsed, gpu.MemoryUsed, 0.01, "Memory used should match")
-				assert.InDelta(t, expected.memoryTotal, gpu.MemoryTotal, 0.01, "Memory total should match")
-				assert.InDelta(t, expected.usage, gpu.Usage, 0.01, "Usage should match")
-				assert.InDelta(t, expected.power, gpu.Power, 0.01, "Power should match")
-				assert.Equal(t, expected.count, gpu.Count, "Count should match")
-			}
-
-			// Verify average calculation in GetCurrentData
-			result := gm.GetCurrentData()
-			for id, expected := range tt.expectedValues {
-				gpu, exists := result[id]
-				assert.True(t, exists, "GPU with ID %s should exist in GetCurrentData result", id)
-				if !exists {
-					continue
-				}
-
-				assert.InDelta(t, expected.temperature, gpu.Temperature, 0.01, "Temperature in GetCurrentData should match")
-				assert.InDelta(t, expected.avgUsage, gpu.Usage, 0.01, "Average usage in GetCurrentData should match")
-				assert.InDelta(t, expected.avgPower, gpu.Power, 0.01, "Average power in GetCurrentData should match")
-			}
-
-			// Verify that accumulators in the original map are reset
-			for id := range tt.expectedValues {
-				gpu, exists := gm.GpuDataMap[id]
-				assert.True(t, exists, "GPU with ID %s should still exist after GetCurrentData", id)
-				if !exists {
-					continue
-				}
-				assert.Equal(t, float64(0), gpu.Count, "Count should be reset for GPU ID %s", id)
-				assert.Equal(t, float64(0), gpu.Usage, "Usage should be reset for GPU ID %s", id)
-				assert.Equal(t, float64(0), gpu.Power, "Power should be reset for GPU ID %s", id)
-			}
-		})
-	}
-}

beszel/internal/agent/network.go

@@ -1,67 +0,0 @@
-package agent
-
-import (
-	"log/slog"
-	"strings"
-	"time"
-
-	psutilNet "github.com/shirou/gopsutil/v4/net"
-)
-
-func (a *Agent) initializeNetIoStats() {
-	// reset valid network interfaces
-	a.netInterfaces = make(map[string]struct{}, 0)
-
-	// map of network interface names passed in via NICS env var
-	var nicsMap map[string]struct{}
-	nics, nicsEnvExists := GetEnv("NICS")
-	if nicsEnvExists {
-		nicsMap = make(map[string]struct{}, 0)
-		for nic := range strings.SplitSeq(nics, ",") {
-			nicsMap[nic] = struct{}{}
-		}
-	}
-
-	// reset network I/O stats
-	a.netIoStats.BytesSent = 0
-	a.netIoStats.BytesRecv = 0
-
-	// get intial network I/O stats
-	if netIO, err := psutilNet.IOCounters(true); err == nil {
-		a.netIoStats.Time = time.Now()
-		for _, v := range netIO {
-			switch {
-			// skip if nics exists and the interface is not in the list
-			case nicsEnvExists:
-				if _, nameInNics := nicsMap[v.Name]; !nameInNics {
-					continue
-				}
-			// otherwise run the interface name through the skipNetworkInterface function
-			default:
-				if a.skipNetworkInterface(v) {
-					continue
-				}
-			}
-			slog.Info("Detected network interface", "name", v.Name, "sent", v.BytesSent, "recv", v.BytesRecv)
-			a.netIoStats.BytesSent += v.BytesSent
-			a.netIoStats.BytesRecv += v.BytesRecv
-			// store as a valid network interface
-			a.netInterfaces[v.Name] = struct{}{}
-		}
-	}
-}
-
-func (a *Agent) skipNetworkInterface(v psutilNet.IOCountersStat) bool {
-	switch {
-	case strings.HasPrefix(v.Name, "lo"),
-		strings.HasPrefix(v.Name, "docker"),
-		strings.HasPrefix(v.Name, "br-"),
-		strings.HasPrefix(v.Name, "veth"),
-		strings.HasPrefix(v.Name, "bond"),
-		v.BytesRecv == 0,
-		v.BytesSent == 0:
-		return true
-	default:
-		return false
-	}
-}

beszel/internal/agent/system.go

@@ -1,304 +0,0 @@
-package agent
-
-import (
-	"beszel"
-	"beszel/internal/agent/battery"
-	"beszel/internal/entities/system"
-	"bufio"
-	"fmt"
-	"log/slog"
-	"os"
-	"strconv"
-	"strings"
-	"time"
-
-	"github.com/shirou/gopsutil/v4/cpu"
-	"github.com/shirou/gopsutil/v4/disk"
-	"github.com/shirou/gopsutil/v4/host"
-	"github.com/shirou/gopsutil/v4/load"
-	"github.com/shirou/gopsutil/v4/mem"
-	psutilNet "github.com/shirou/gopsutil/v4/net"
-)
-
-// Sets initial / non-changing values about the host system
-func (a *Agent) initializeSystemInfo() {
-	a.systemInfo.AgentVersion = beszel.Version
-	a.systemInfo.Hostname, _ = os.Hostname()
-
-	platform, _, version, _ := host.PlatformInformation()
-
-	if platform == "darwin" {
-		a.systemInfo.KernelVersion = version
-		a.systemInfo.Os = system.Darwin
-	} else if strings.Contains(platform, "indows") {
-		a.systemInfo.KernelVersion = strings.Replace(platform, "Microsoft ", "", 1) + " " + version
-		a.systemInfo.Os = system.Windows
-	} else if platform == "freebsd" {
-		a.systemInfo.Os = system.Freebsd
-		a.systemInfo.KernelVersion = version
-	} else {
-		a.systemInfo.Os = system.Linux
-	}
-
-	if a.systemInfo.KernelVersion == "" {
-		a.systemInfo.KernelVersion, _ = host.KernelVersion()
-	}
-
-	// cpu model
-	if info, err := cpu.Info(); err == nil && len(info) > 0 {
-		a.systemInfo.CpuModel = info[0].ModelName
-	}
-	// cores / threads
-	a.systemInfo.Cores, _ = cpu.Counts(false)
-	if threads, err := cpu.Counts(true); err == nil {
-		if threads > 0 && threads < a.systemInfo.Cores {
-			// in lxc logical cores reflects container limits, so use that as cores if lower
-			a.systemInfo.Cores = threads
-		} else {
-			a.systemInfo.Threads = threads
-		}
-	}
-
-	// zfs
-	if _, err := getARCSize(); err != nil {
-		slog.Debug("Not monitoring ZFS ARC", "err", err)
-	} else {
-		a.zfs = true
-	}
-}
-
-// Returns current info, stats about the host system
-func (a *Agent) getSystemStats() system.Stats {
-	systemStats := system.Stats{}
-
-	// battery
-	if battery.HasReadableBattery() {
-		systemStats.Battery[0], systemStats.Battery[1], _ = battery.GetBatteryStats()
-	}
-
-	// cpu percent
-	cpuPct, err := cpu.Percent(0, false)
-	if err != nil {
-		slog.Error("Error getting cpu percent", "err", err)
-	} else if len(cpuPct) > 0 {
-		systemStats.Cpu = twoDecimals(cpuPct[0])
-	}
-
-	// load average
-	if avgstat, err := load.Avg(); err == nil {
-		systemStats.LoadAvg[0] = avgstat.Load1
-		systemStats.LoadAvg[1] = avgstat.Load5
-		systemStats.LoadAvg[2] = avgstat.Load15
-		slog.Debug("Load average", "5m", avgstat.Load5, "15m", avgstat.Load15)
-	} else {
-		slog.Error("Error getting load average", "err", err)
-	}
-
-	// memory
-	if v, err := mem.VirtualMemory(); err == nil {
-		// swap
-		systemStats.Swap = bytesToGigabytes(v.SwapTotal)
-		systemStats.SwapUsed = bytesToGigabytes(v.SwapTotal - v.SwapFree - v.SwapCached)
-		// cache + buffers value for default mem calculation
-		cacheBuff := v.Total - v.Free - v.Used
-		// htop memory calculation overrides
-		if a.memCalc == "htop" {
-			// note: gopsutil automatically adds SReclaimable to v.Cached
-			cacheBuff = v.Cached + v.Buffers - v.Shared
-			v.Used = v.Total - (v.Free + cacheBuff)
-			v.UsedPercent = float64(v.Used) / float64(v.Total) * 100.0
-		}
-		// subtract ZFS ARC size from used memory and add as its own category
-		if a.zfs {
-			if arcSize, _ := getARCSize(); arcSize > 0 && arcSize < v.Used {
-				v.Used = v.Used - arcSize
-				v.UsedPercent = float64(v.Used) / float64(v.Total) * 100.0
-				systemStats.MemZfsArc = bytesToGigabytes(arcSize)
-			}
-		}
-		systemStats.Mem = bytesToGigabytes(v.Total)
-		systemStats.MemBuffCache = bytesToGigabytes(cacheBuff)
-		systemStats.MemUsed = bytesToGigabytes(v.Used)
-		systemStats.MemPct = twoDecimals(v.UsedPercent)
-	}
-
-	// disk usage
-	for _, stats := range a.fsStats {
-		if d, err := disk.Usage(stats.Mountpoint); err == nil {
-			stats.DiskTotal = bytesToGigabytes(d.Total)
-			stats.DiskUsed = bytesToGigabytes(d.Used)
-			if stats.Root {
-				systemStats.DiskTotal = bytesToGigabytes(d.Total)
-				systemStats.DiskUsed = bytesToGigabytes(d.Used)
-				systemStats.DiskPct = twoDecimals(d.UsedPercent)
-			}
-		} else {
-			// reset stats if error (likely unmounted)
-			slog.Error("Error getting disk stats", "name", stats.Mountpoint, "err", err)
-			stats.DiskTotal = 0
-			stats.DiskUsed = 0
-			stats.TotalRead = 0
-			stats.TotalWrite = 0
-		}
-	}
-
-	// disk i/o
-	if ioCounters, err := disk.IOCounters(a.fsNames...); err == nil {
-		for _, d := range ioCounters {
-			stats := a.fsStats[d.Name]
-			if stats == nil {
-				continue
-			}
-			secondsElapsed := time.Since(stats.Time).Seconds()
-			readPerSecond := bytesToMegabytes(float64(d.ReadBytes-stats.TotalRead) / secondsElapsed)
-			writePerSecond := bytesToMegabytes(float64(d.WriteBytes-stats.TotalWrite) / secondsElapsed)
-			// check for invalid values and reset stats if so
-			if readPerSecond < 0 || writePerSecond < 0 || readPerSecond > 50_000 || writePerSecond > 50_000 {
-				slog.Warn("Invalid disk I/O. Resetting.", "name", d.Name, "read", readPerSecond, "write", writePerSecond)
-				a.initializeDiskIoStats(ioCounters)
-				break
-			}
-			stats.Time = time.Now()
-			stats.DiskReadPs = readPerSecond
-			stats.DiskWritePs = writePerSecond
-			stats.TotalRead = d.ReadBytes
-			stats.TotalWrite = d.WriteBytes
-			// if root filesystem, update system stats
-			if stats.Root {
-				systemStats.DiskReadPs = stats.DiskReadPs
-				systemStats.DiskWritePs = stats.DiskWritePs
-			}
-		}
-	}
-
-	// network stats
-	if len(a.netInterfaces) == 0 {
-		// if no network interfaces, initialize again
-		// this is a fix if agent started before network is online (#466)
-		// maybe refactor this in the future to not cache interface names at all so we
-		// don't miss an interface that's been added after agent started in any circumstance
-		a.initializeNetIoStats()
-	}
-	if netIO, err := psutilNet.IOCounters(true); err == nil {
-		msElapsed := uint64(time.Since(a.netIoStats.Time).Milliseconds())
-		a.netIoStats.Time = time.Now()
-		totalBytesSent := uint64(0)
-		totalBytesRecv := uint64(0)
-		// sum all bytes sent and received
-		for _, v := range netIO {
-			// skip if not in valid network interfaces list
-			if _, exists := a.netInterfaces[v.Name]; !exists {
-				continue
-			}
-			totalBytesSent += v.BytesSent
-			totalBytesRecv += v.BytesRecv
-		}
-		// add to systemStats
-		var bytesSentPerSecond, bytesRecvPerSecond uint64
-		if msElapsed > 0 {
-			bytesSentPerSecond = (totalBytesSent - a.netIoStats.BytesSent) * 1000 / msElapsed
-			bytesRecvPerSecond = (totalBytesRecv - a.netIoStats.BytesRecv) * 1000 / msElapsed
-		}
-		networkSentPs := bytesToMegabytes(float64(bytesSentPerSecond))
-		networkRecvPs := bytesToMegabytes(float64(bytesRecvPerSecond))
-		// add check for issue (#150) where sent is a massive number
-		if networkSentPs > 10_000 || networkRecvPs > 10_000 {
-			slog.Warn("Invalid net stats. Resetting.", "sent", networkSentPs, "recv", networkRecvPs)
-			for _, v := range netIO {
-				if _, exists := a.netInterfaces[v.Name]; !exists {
-					continue
-				}
-				slog.Info(v.Name, "recv", v.BytesRecv, "sent", v.BytesSent)
-			}
-			// reset network I/O stats
-			a.initializeNetIoStats()
-		} else {
-			systemStats.NetworkSent = networkSentPs
-			systemStats.NetworkRecv = networkRecvPs
-			systemStats.Bandwidth[0], systemStats.Bandwidth[1] = bytesSentPerSecond, bytesRecvPerSecond
-			// update netIoStats
-			a.netIoStats.BytesSent = totalBytesSent
-			a.netIoStats.BytesRecv = totalBytesRecv
-		}
-	}
-
-	// temperatures
-	// TODO: maybe refactor to methods on systemStats
-	a.updateTemperatures(&systemStats)
-
-	// GPU data
-	if a.gpuManager != nil {
-		// reset high gpu percent
-		a.systemInfo.GpuPct = 0
-		// get current GPU data
-		if gpuData := a.gpuManager.GetCurrentData(); len(gpuData) > 0 {
-			systemStats.GPUData = gpuData
-
-			// add temperatures
-			if systemStats.Temperatures == nil {
-				systemStats.Temperatures = make(map[string]float64, len(gpuData))
-			}
-			highestTemp := 0.0
-			for _, gpu := range gpuData {
-				if gpu.Temperature > 0 {
-					systemStats.Temperatures[gpu.Name] = gpu.Temperature
-					if a.sensorConfig.primarySensor == gpu.Name {
-						a.systemInfo.DashboardTemp = gpu.Temperature
-					}
-					if gpu.Temperature > highestTemp {
-						highestTemp = gpu.Temperature
-					}
-				}
-				// update high gpu percent for dashboard
-				a.systemInfo.GpuPct = max(a.systemInfo.GpuPct, gpu.Usage)
-			}
-			// use highest temp for dashboard temp if dashboard temp is unset
-			if a.systemInfo.DashboardTemp == 0 {
-				a.systemInfo.DashboardTemp = highestTemp
-			}
-		}
-	}
-
-	// update base system info
-	a.systemInfo.Cpu = systemStats.Cpu
-	a.systemInfo.LoadAvg = systemStats.LoadAvg
-	// TODO: remove these in future release in favor of load avg array
-	a.systemInfo.LoadAvg1 = systemStats.LoadAvg[0]
-	a.systemInfo.LoadAvg5 = systemStats.LoadAvg[1]
-	a.systemInfo.LoadAvg15 = systemStats.LoadAvg[2]
-	a.systemInfo.MemPct = systemStats.MemPct
-	a.systemInfo.DiskPct = systemStats.DiskPct
-	a.systemInfo.Uptime, _ = host.Uptime()
-	// TODO: in future release, remove MB bandwidth values in favor of bytes
-	a.systemInfo.Bandwidth = twoDecimals(systemStats.NetworkSent + systemStats.NetworkRecv)
-	a.systemInfo.BandwidthBytes = systemStats.Bandwidth[0] + systemStats.Bandwidth[1]
-	slog.Debug("sysinfo", "data", a.systemInfo)
-
-	return systemStats
-}
-
-// Returns the size of the ZFS ARC memory cache in bytes
-func getARCSize() (uint64, error) {
-	file, err := os.Open("/proc/spl/kstat/zfs/arcstats")
-	if err != nil {
-		return 0, err
-	}
-	defer file.Close()
-
-	// Scan the lines
-	scanner := bufio.NewScanner(file)
-	for scanner.Scan() {
-		line := scanner.Text()
-		if strings.HasPrefix(line, "size") {
-			// Example line: size 4 15032385536
-			fields := strings.Fields(line)
-			if len(fields) < 3 {
-				return 0, err
-			}
-			// Return the size as uint64
-			return strconv.ParseUint(fields[2], 10, 64)
-		}
-	}
-
-	return 0, fmt.Errorf("failed to parse size field")
-}

beszel/internal/alerts/alerts_test.go

@@ -1,368 +0,0 @@
-//go:build testing
-// +build testing
-
-package alerts_test
-
-import (
-	"bytes"
-	"encoding/json"
-	"io"
-	"net/http"
-	"strings"
-	"testing"
-
-	beszelTests "beszel/internal/tests"
-
-	"github.com/pocketbase/dbx"
-	"github.com/pocketbase/pocketbase/core"
-	pbTests "github.com/pocketbase/pocketbase/tests"
-	"github.com/stretchr/testify/assert"
-)
-
-// marshal to json and return an io.Reader (for use in ApiScenario.Body)
-func jsonReader(v any) io.Reader {
-	data, err := json.Marshal(v)
-	if err != nil {
-		panic(err)
-	}
-	return bytes.NewReader(data)
-}
-
-func TestUserAlertsApi(t *testing.T) {
-	hub, _ := beszelTests.NewTestHub(t.TempDir())
-	defer hub.Cleanup()
-
-	hub.StartHub()
-
-	user1, _ := beszelTests.CreateUser(hub, "alertstest@example.com", "password")
-	user1Token, _ := user1.NewAuthToken()
-
-	user2, _ := beszelTests.CreateUser(hub, "alertstest2@example.com", "password")
-	user2Token, _ := user2.NewAuthToken()
-
-	system1, _ := beszelTests.CreateRecord(hub, "systems", map[string]any{
-		"name":  "system1",
-		"users": []string{user1.Id},
-		"host":  "127.0.0.1",
-	})
-
-	system2, _ := beszelTests.CreateRecord(hub, "systems", map[string]any{
-		"name":  "system2",
-		"users": []string{user1.Id, user2.Id},
-		"host":  "127.0.0.2",
-	})
-
-	userRecords, _ := hub.CountRecords("users")
-	assert.EqualValues(t, 2, userRecords, "all users should be created")
-
-	systemRecords, _ := hub.CountRecords("systems")
-	assert.EqualValues(t, 2, systemRecords, "all systems should be created")
-
-	testAppFactory := func(t testing.TB) *pbTests.TestApp {
-		return hub.TestApp
-	}
-
-	scenarios := []beszelTests.ApiScenario{
-		{
-			Name:            "GET not implemented - returns index",
-			Method:          http.MethodGet,
-			URL:             "/api/beszel/user-alerts",
-			ExpectedStatus:  200,
-			ExpectedContent: []string{"<html ", "globalThis.BESZEL"},
-			TestAppFactory:  testAppFactory,
-		},
-		{
-			Name:            "POST no auth",
-			Method:          http.MethodPost,
-			URL:             "/api/beszel/user-alerts",
-			ExpectedStatus:  401,
-			ExpectedContent: []string{"requires valid"},
-			TestAppFactory:  testAppFactory,
-		},
-		{
-			Name:   "POST no body",
-			Method: http.MethodPost,
-			URL:    "/api/beszel/user-alerts",
-			Headers: map[string]string{
-				"Authorization": user1Token,
-			},
-			ExpectedStatus:  400,
-			ExpectedContent: []string{"Bad data"},
-			TestAppFactory:  testAppFactory,
-		},
-		{
-			Name:   "POST bad data",
-			Method: http.MethodPost,
-			URL:    "/api/beszel/user-alerts",
-			Headers: map[string]string{
-				"Authorization": user1Token,
-			},
-			ExpectedStatus:  400,
-			ExpectedContent: []string{"Bad data"},
-			TestAppFactory:  testAppFactory,
-			Body: jsonReader(map[string]any{
-				"invalidField": "this should cause validation error",
-				"threshold":    "not a number",
-			}),
-		},
-		{
-			Name:   "POST malformed JSON",
-			Method: http.MethodPost,
-			URL:    "/api/beszel/user-alerts",
-			Headers: map[string]string{
-				"Authorization": user1Token,
-			},
-			ExpectedStatus:  400,
-			ExpectedContent: []string{"Bad data"},
-			TestAppFactory:  testAppFactory,
-			Body:            strings.NewReader(`{"alertType": "cpu", "threshold": 80, "enabled": true,}`),
-		},
-		{
-			Name:   "POST valid alert data multiple systems",
-			Method: http.MethodPost,
-			URL:    "/api/beszel/user-alerts",
-			Headers: map[string]string{
-				"Authorization": user1Token,
-			},
-			ExpectedStatus:  200,
-			ExpectedContent: []string{"\"success\":true"},
-			TestAppFactory:  testAppFactory,
-			Body: jsonReader(map[string]any{
-				"name":      "CPU",
-				"value":     69,
-				"min":       9,
-				"systems":   []string{system1.Id, system2.Id},
-				"overwrite": false,
-			}),
-			AfterTestFunc: func(t testing.TB, app *pbTests.TestApp, res *http.Response) {
-				// check total alerts
-				alerts, _ := app.CountRecords("alerts")
-				assert.EqualValues(t, 2, alerts, "should have 2 alerts")
-				// check alert has correct values
-				matchingAlerts, _ := app.CountRecords("alerts", dbx.HashExp{"name": "CPU", "user": user1.Id, "system": system1.Id, "value": 69, "min": 9})
-				assert.EqualValues(t, 1, matchingAlerts, "should have 1 alert")
-			},
-		},
-		{
-			Name:   "POST valid alert data single system",
-			Method: http.MethodPost,
-			URL:    "/api/beszel/user-alerts",
-			Headers: map[string]string{
-				"Authorization": user1Token,
-			},
-			ExpectedStatus:  200,
-			ExpectedContent: []string{"\"success\":true"},
-			TestAppFactory:  testAppFactory,
-			Body: jsonReader(map[string]any{
-				"name":    "Memory",
-				"systems": []string{system1.Id},
-				"value":   90,
-				"min":     10,
-			}),
-			AfterTestFunc: func(t testing.TB, app *pbTests.TestApp, res *http.Response) {
-				user1Alerts, _ := app.CountRecords("alerts", dbx.HashExp{"user": user1.Id})
-				assert.EqualValues(t, 3, user1Alerts, "should have 3 alerts")
-			},
-		},
-		{
-			Name:   "Overwrite: false, should not overwrite existing alert",
-			Method: http.MethodPost,
-			URL:    "/api/beszel/user-alerts",
-			Headers: map[string]string{
-				"Authorization": user1Token,
-			},
-			ExpectedStatus:  200,
-			ExpectedContent: []string{"\"success\":true"},
-			TestAppFactory:  testAppFactory,
-			Body: jsonReader(map[string]any{
-				"name":      "CPU",
-				"value":     45,
-				"min":       5,
-				"systems":   []string{system1.Id},
-				"overwrite": false,
-			}),
-			BeforeTestFunc: func(t testing.TB, app *pbTests.TestApp, e *core.ServeEvent) {
-				beszelTests.ClearCollection(t, app, "alerts")
-				beszelTests.CreateRecord(app, "alerts", map[string]any{
-					"name":   "CPU",
-					"system": system1.Id,
-					"user":   user1.Id,
-					"value":  80,
-					"min":    10,
-				})
-			},
-			AfterTestFunc: func(t testing.TB, app *pbTests.TestApp, res *http.Response) {
-				alerts, _ := app.CountRecords("alerts")
-				assert.EqualValues(t, 1, alerts, "should have 1 alert")
-				alert, _ := app.FindFirstRecordByFilter("alerts", "name = 'CPU' && user = {:user}", dbx.Params{"user": user1.Id})
-				assert.EqualValues(t, 80, alert.Get("value"), "should have 80 as value")
-			},
-		},
-		{
-			Name:   "Overwrite: true, should overwrite existing alert",
-			Method: http.MethodPost,
-			URL:    "/api/beszel/user-alerts",
-			Headers: map[string]string{
-				"Authorization": user2Token,
-			},
-			ExpectedStatus:  200,
-			ExpectedContent: []string{"\"success\":true"},
-			TestAppFactory:  testAppFactory,
-			Body: jsonReader(map[string]any{
-				"name":      "CPU",
-				"value":     45,
-				"min":       5,
-				"systems":   []string{system2.Id},
-				"overwrite": true,
-			}),
-			BeforeTestFunc: func(t testing.TB, app *pbTests.TestApp, e *core.ServeEvent) {
-				beszelTests.ClearCollection(t, app, "alerts")
-				beszelTests.CreateRecord(app, "alerts", map[string]any{
-					"name":   "CPU",
-					"system": system2.Id,
-					"user":   user2.Id,
-					"value":  80,
-					"min":    10,
-				})
-			},
-			AfterTestFunc: func(t testing.TB, app *pbTests.TestApp, res *http.Response) {
-				alerts, _ := app.CountRecords("alerts")
-				assert.EqualValues(t, 1, alerts, "should have 1 alert")
-				alert, _ := app.FindFirstRecordByFilter("alerts", "name = 'CPU' && user = {:user}", dbx.Params{"user": user2.Id})
-				assert.EqualValues(t, 45, alert.Get("value"), "should have 45 as value")
-			},
-		},
-		{
-			Name:            "DELETE no auth",
-			Method:          http.MethodDelete,
-			URL:             "/api/beszel/user-alerts",
-			ExpectedStatus:  401,
-			ExpectedContent: []string{"requires valid"},
-			TestAppFactory:  testAppFactory,
-			Body: jsonReader(map[string]any{
-				"name":    "CPU",
-				"systems": []string{system1.Id},
-			}),
-			BeforeTestFunc: func(t testing.TB, app *pbTests.TestApp, e *core.ServeEvent) {
-				beszelTests.ClearCollection(t, app, "alerts")
-				beszelTests.CreateRecord(app, "alerts", map[string]any{
-					"name":   "CPU",
-					"system": system1.Id,
-					"user":   user1.Id,
-					"value":  80,
-					"min":    10,
-				})
-			},
-			AfterTestFunc: func(t testing.TB, app *pbTests.TestApp, res *http.Response) {
-				alerts, _ := app.CountRecords("alerts")
-				assert.EqualValues(t, 1, alerts, "should have 1 alert")
-			},
-		},
-		{
-			Name:   "DELETE alert",
-			Method: http.MethodDelete,
-			URL:    "/api/beszel/user-alerts",
-			Headers: map[string]string{
-				"Authorization": user1Token,
-			},
-			ExpectedStatus:  200,
-			ExpectedContent: []string{"\"count\":1", "\"success\":true"},
-			TestAppFactory:  testAppFactory,
-			Body: jsonReader(map[string]any{
-				"name":    "CPU",
-				"systems": []string{system1.Id},
-			}),
-			BeforeTestFunc: func(t testing.TB, app *pbTests.TestApp, e *core.ServeEvent) {
-				beszelTests.ClearCollection(t, app, "alerts")
-				beszelTests.CreateRecord(app, "alerts", map[string]any{
-					"name":   "CPU",
-					"system": system1.Id,
-					"user":   user1.Id,
-					"value":  80,
-					"min":    10,
-				})
-			},
-			AfterTestFunc: func(t testing.TB, app *pbTests.TestApp, res *http.Response) {
-				alerts, _ := app.CountRecords("alerts")
-				assert.Zero(t, alerts, "should have 0 alerts")
-			},
-		},
-		{
-			Name:   "DELETE alert multiple systems",
-			Method: http.MethodDelete,
-			URL:    "/api/beszel/user-alerts",
-			Headers: map[string]string{
-				"Authorization": user1Token,
-			},
-			ExpectedStatus:  200,
-			ExpectedContent: []string{"\"count\":2", "\"success\":true"},
-			TestAppFactory:  testAppFactory,
-			Body: jsonReader(map[string]any{
-				"name":    "Memory",
-				"systems": []string{system1.Id, system2.Id},
-			}),
-			BeforeTestFunc: func(t testing.TB, app *pbTests.TestApp, e *core.ServeEvent) {
-				beszelTests.ClearCollection(t, app, "alerts")
-				for _, systemId := range []string{system1.Id, system2.Id} {
-					_, err := beszelTests.CreateRecord(app, "alerts", map[string]any{
-						"name":   "Memory",
-						"system": systemId,
-						"user":   user1.Id,
-						"value":  90,
-						"min":    10,
-					})
-					assert.NoError(t, err, "should create alert")
-				}
-				alerts, _ := app.CountRecords("alerts")
-				assert.EqualValues(t, 2, alerts, "should have 2 alerts")
-			},
-			AfterTestFunc: func(t testing.TB, app *pbTests.TestApp, res *http.Response) {
-				alerts, _ := app.CountRecords("alerts")
-				assert.Zero(t, alerts, "should have 0 alerts")
-			},
-		},
-		{
-			Name:   "User 2 should not be able to delete alert of user 1",
-			Method: http.MethodDelete,
-			URL:    "/api/beszel/user-alerts",
-			Headers: map[string]string{
-				"Authorization": user2Token,
-			},
-			ExpectedStatus:  200,
-			ExpectedContent: []string{"\"count\":1", "\"success\":true"},
-			TestAppFactory:  testAppFactory,
-			Body: jsonReader(map[string]any{
-				"name":    "CPU",
-				"systems": []string{system2.Id},
-			}),
-			BeforeTestFunc: func(t testing.TB, app *pbTests.TestApp, e *core.ServeEvent) {
-				beszelTests.ClearCollection(t, app, "alerts")
-				for _, user := range []string{user1.Id, user2.Id} {
-					beszelTests.CreateRecord(app, "alerts", map[string]any{
-						"name":   "CPU",
-						"system": system2.Id,
-						"user":   user,
-						"value":  80,
-						"min":    10,
-					})
-				}
-				alerts, _ := app.CountRecords("alerts")
-				assert.EqualValues(t, 2, alerts, "should have 2 alerts")
-				user1AlertCount, _ := app.CountRecords("alerts", dbx.HashExp{"user": user1.Id})
-				assert.EqualValues(t, 1, user1AlertCount, "should have 1 alert")
-				user2AlertCount, _ := app.CountRecords("alerts", dbx.HashExp{"user": user2.Id})
-				assert.EqualValues(t, 1, user2AlertCount, "should have 1 alert")
-			},
-			AfterTestFunc: func(t testing.TB, app *pbTests.TestApp, res *http.Response) {
-				user1AlertCount, _ := app.CountRecords("alerts", dbx.HashExp{"user": user1.Id})
-				assert.EqualValues(t, 1, user1AlertCount, "should have 1 alert")
-				user2AlertCount, _ := app.CountRecords("alerts", dbx.HashExp{"user": user2.Id})
-				assert.Zero(t, user2AlertCount, "should have 0 alerts")
-			},
-		},
-	}
-
-	for _, scenario := range scenarios {
-		scenario.Test(t)
-	}
-}

beszel/internal/common/common-ws.go

@@ -1,32 +0,0 @@
-package common
-
-type WebSocketAction = uint8
-
-// Not implemented yet
-// type AgentError = uint8
-
-const (
-	// Request system data from agent
-	GetData WebSocketAction = iota
-	// Check the fingerprint of the agent
-	CheckFingerprint
-)
-
-// HubRequest defines the structure for requests sent from hub to agent.
-type HubRequest[T any] struct {
-	Action WebSocketAction `cbor:"0,keyasint"`
-	Data   T               `cbor:"1,keyasint,omitempty,omitzero"`
-	// Error  AgentError      `cbor:"error,omitempty,omitzero"`
-}
-
-type FingerprintRequest struct {
-	Signature   []byte `cbor:"0,keyasint"`
-	NeedSysInfo bool   `cbor:"1,keyasint"` // For universal token system creation
-}
-
-type FingerprintResponse struct {
-	Fingerprint string `cbor:"0,keyasint"`
-	// Optional system info for universal token system creation
-	Hostname string `cbor:"1,keyasint,omitempty,omitzero"`
-	Port     string `cbor:"2,keyasint,omitempty,omitzero"`
-}

beszel/internal/entities/system/system.go

@@ -1,114 +0,0 @@
-package system
-
-// TODO: this is confusing, make common package with common/types common/helpers etc
-
-import (
-	"beszel/internal/entities/container"
-	"time"
-)
-
-type Stats struct {
-	Cpu            float64             `json:"cpu" cbor:"0,keyasint"`
-	MaxCpu         float64             `json:"cpum,omitempty" cbor:"1,keyasint,omitempty"`
-	Mem            float64             `json:"m" cbor:"2,keyasint"`
-	MemUsed        float64             `json:"mu" cbor:"3,keyasint"`
-	MemPct         float64             `json:"mp" cbor:"4,keyasint"`
-	MemBuffCache   float64             `json:"mb" cbor:"5,keyasint"`
-	MemZfsArc      float64             `json:"mz,omitempty" cbor:"6,keyasint,omitempty"` // ZFS ARC memory
-	Swap           float64             `json:"s,omitempty" cbor:"7,keyasint,omitempty"`
-	SwapUsed       float64             `json:"su,omitempty" cbor:"8,keyasint,omitempty"`
-	DiskTotal      float64             `json:"d" cbor:"9,keyasint"`
-	DiskUsed       float64             `json:"du" cbor:"10,keyasint"`
-	DiskPct        float64             `json:"dp" cbor:"11,keyasint"`
-	DiskReadPs     float64             `json:"dr" cbor:"12,keyasint"`
-	DiskWritePs    float64             `json:"dw" cbor:"13,keyasint"`
-	MaxDiskReadPs  float64             `json:"drm,omitempty" cbor:"14,keyasint,omitempty"`
-	MaxDiskWritePs float64             `json:"dwm,omitempty" cbor:"15,keyasint,omitempty"`
-	NetworkSent    float64             `json:"ns" cbor:"16,keyasint"`
-	NetworkRecv    float64             `json:"nr" cbor:"17,keyasint"`
-	MaxNetworkSent float64             `json:"nsm,omitempty" cbor:"18,keyasint,omitempty"`
-	MaxNetworkRecv float64             `json:"nrm,omitempty" cbor:"19,keyasint,omitempty"`
-	Temperatures   map[string]float64  `json:"t,omitempty" cbor:"20,keyasint,omitempty"`
-	ExtraFs        map[string]*FsStats `json:"efs,omitempty" cbor:"21,keyasint,omitempty"`
-	GPUData        map[string]GPUData  `json:"g,omitempty" cbor:"22,keyasint,omitempty"`
-	LoadAvg1       float64             `json:"l1,omitempty" cbor:"23,keyasint,omitempty"`
-	LoadAvg5       float64             `json:"l5,omitempty" cbor:"24,keyasint,omitempty"`
-	LoadAvg15      float64             `json:"l15,omitempty" cbor:"25,keyasint,omitempty"`
-	Bandwidth      [2]uint64           `json:"b,omitzero" cbor:"26,keyasint,omitzero"`  // [sent bytes, recv bytes]
-	MaxBandwidth   [2]uint64           `json:"bm,omitzero" cbor:"27,keyasint,omitzero"` // [sent bytes, recv bytes]
-	// TODO: remove other load fields in future release in favor of load avg array
-	LoadAvg [3]float64 `json:"la,omitempty" cbor:"28,keyasint"`
-	Battery [2]uint8   `json:"bat,omitzero" cbor:"29,keyasint,omitzero"` // [percent, charge state, current]
-	MaxMem  float64    `json:"mm,omitempty" cbor:"30,keyasint,omitempty"`
-}
-
-type GPUData struct {
-	Name        string  `json:"n" cbor:"0,keyasint"`
-	Temperature float64 `json:"-"`
-	MemoryUsed  float64 `json:"mu,omitempty" cbor:"1,keyasint,omitempty"`
-	MemoryTotal float64 `json:"mt,omitempty" cbor:"2,keyasint,omitempty"`
-	Usage       float64 `json:"u" cbor:"3,keyasint"`
-	Power       float64 `json:"p,omitempty" cbor:"4,keyasint,omitempty"`
-	Count       float64 `json:"-"`
-}
-
-type FsStats struct {
-	Time           time.Time `json:"-"`
-	Root           bool      `json:"-"`
-	Mountpoint     string    `json:"-"`
-	DiskTotal      float64   `json:"d" cbor:"0,keyasint"`
-	DiskUsed       float64   `json:"du" cbor:"1,keyasint"`
-	TotalRead      uint64    `json:"-"`
-	TotalWrite     uint64    `json:"-"`
-	DiskReadPs     float64   `json:"r" cbor:"2,keyasint"`
-	DiskWritePs    float64   `json:"w" cbor:"3,keyasint"`
-	MaxDiskReadPS  float64   `json:"rm,omitempty" cbor:"4,keyasint,omitempty"`
-	MaxDiskWritePS float64   `json:"wm,omitempty" cbor:"5,keyasint,omitempty"`
-}
-
-type NetIoStats struct {
-	BytesRecv uint64
-	BytesSent uint64
-	Time      time.Time
-	Name      string
-}
-
-type Os = uint8
-
-const (
-	Linux Os = iota
-	Darwin
-	Windows
-	Freebsd
-)
-
-type Info struct {
-	Hostname       string  `json:"h" cbor:"0,keyasint"`
-	KernelVersion  string  `json:"k,omitempty" cbor:"1,keyasint,omitempty"`
-	Cores          int     `json:"c" cbor:"2,keyasint"`
-	Threads        int     `json:"t,omitempty" cbor:"3,keyasint,omitempty"`
-	CpuModel       string  `json:"m" cbor:"4,keyasint"`
-	Uptime         uint64  `json:"u" cbor:"5,keyasint"`
-	Cpu            float64 `json:"cpu" cbor:"6,keyasint"`
-	MemPct         float64 `json:"mp" cbor:"7,keyasint"`
-	DiskPct        float64 `json:"dp" cbor:"8,keyasint"`
-	Bandwidth      float64 `json:"b" cbor:"9,keyasint"`
-	AgentVersion   string  `json:"v" cbor:"10,keyasint"`
-	Podman         bool    `json:"p,omitempty" cbor:"11,keyasint,omitempty"`
-	GpuPct         float64 `json:"g,omitempty" cbor:"12,keyasint,omitempty"`
-	DashboardTemp  float64 `json:"dt,omitempty" cbor:"13,keyasint,omitempty"`
-	Os             Os      `json:"os" cbor:"14,keyasint"`
-	LoadAvg1       float64 `json:"l1,omitempty" cbor:"15,keyasint,omitempty"`
-	LoadAvg5       float64 `json:"l5,omitempty" cbor:"16,keyasint,omitempty"`
-	LoadAvg15      float64 `json:"l15,omitempty" cbor:"17,keyasint,omitempty"`
-	BandwidthBytes uint64  `json:"bb" cbor:"18,keyasint"`
-	// TODO: remove load fields in future release in favor of load avg array
-	LoadAvg [3]float64 `json:"la,omitempty" cbor:"19,keyasint"`
-}
-
-// Final data structure to return to the hub
-type CombinedData struct {
-	Stats      Stats              `json:"stats" cbor:"0,keyasint"`
-	Info       Info               `json:"info" cbor:"1,keyasint"`
-	Containers []*container.Stats `json:"container" cbor:"2,keyasint"`
-}

beszel/internal/hub/hub.go

@@ -1,351 +0,0 @@
-// Package hub handles updating systems and serving the web UI.
-package hub
-
-import (
-	"beszel"
-	"beszel/internal/alerts"
-	"beszel/internal/hub/config"
-	"beszel/internal/hub/systems"
-	"beszel/internal/records"
-	"beszel/internal/users"
-	"beszel/site"
-	"crypto/ed25519"
-	"encoding/pem"
-	"fmt"
-	"io/fs"
-	"net/http"
-	"net/http/httputil"
-	"net/url"
-	"os"
-	"path"
-	"strings"
-	"time"
-
-	"github.com/google/uuid"
-	"github.com/pocketbase/pocketbase"
-	"github.com/pocketbase/pocketbase/apis"
-	"github.com/pocketbase/pocketbase/core"
-	"golang.org/x/crypto/ssh"
-)
-
-type Hub struct {
-	core.App
-	*alerts.AlertManager
-	um     *users.UserManager
-	rm     *records.RecordManager
-	sm     *systems.SystemManager
-	pubKey string
-	signer ssh.Signer
-	appURL string
-}
-
-// NewHub creates a new Hub instance with default configuration
-func NewHub(app core.App) *Hub {
-	hub := &Hub{}
-	hub.App = app
-
-	hub.AlertManager = alerts.NewAlertManager(hub)
-	hub.um = users.NewUserManager(hub)
-	hub.rm = records.NewRecordManager(hub)
-	hub.sm = systems.NewSystemManager(hub)
-	hub.appURL, _ = GetEnv("APP_URL")
-	return hub
-}
-
-// GetEnv retrieves an environment variable with a "BESZEL_HUB_" prefix, or falls back to the unprefixed key.
-func GetEnv(key string) (value string, exists bool) {
-	if value, exists = os.LookupEnv("BESZEL_HUB_" + key); exists {
-		return value, exists
-	}
-	// Fallback to the old unprefixed key
-	return os.LookupEnv(key)
-}
-
-func (h *Hub) StartHub() error {
-	h.App.OnServe().BindFunc(func(e *core.ServeEvent) error {
-		// initialize settings / collections
-		if err := h.initialize(e); err != nil {
-			return err
-		}
-		// sync systems with config
-		if err := config.SyncSystems(e); err != nil {
-			return err
-		}
-		// register api routes
-		if err := h.registerApiRoutes(e); err != nil {
-			return err
-		}
-		// register cron jobs
-		if err := h.registerCronJobs(e); err != nil {
-			return err
-		}
-		// start server
-		if err := h.startServer(e); err != nil {
-			return err
-		}
-		// start system updates
-		if err := h.sm.Initialize(); err != nil {
-			return err
-		}
-		return e.Next()
-	})
-
-	// TODO: move to users package
-	// handle default values for user / user_settings creation
-	h.App.OnRecordCreate("users").BindFunc(h.um.InitializeUserRole)
-	h.App.OnRecordCreate("user_settings").BindFunc(h.um.InitializeUserSettings)
-
-	if pb, ok := h.App.(*pocketbase.PocketBase); ok {
-		// log.Println("Starting pocketbase")
-		err := pb.Start()
-		if err != nil {
-			return err
-		}
-	}
-
-	return nil
-}
-
-// initialize sets up initial configuration (collections, settings, etc.)
-func (h *Hub) initialize(e *core.ServeEvent) error {
-	// set general settings
-	settings := e.App.Settings()
-	// batch requests (for global alerts)
-	settings.Batch.Enabled = true
-	// set URL if BASE_URL env is set
-	if h.appURL != "" {
-		settings.Meta.AppURL = h.appURL
-	}
-	if err := e.App.Save(settings); err != nil {
-		return err
-	}
-	// set auth settings
-	usersCollection, err := e.App.FindCollectionByNameOrId("users")
-	if err != nil {
-		return err
-	}
-	// disable email auth if DISABLE_PASSWORD_AUTH env var is set
-	disablePasswordAuth, _ := GetEnv("DISABLE_PASSWORD_AUTH")
-	usersCollection.PasswordAuth.Enabled = disablePasswordAuth != "true"
-	usersCollection.PasswordAuth.IdentityFields = []string{"email"}
-	// disable oauth if no providers are configured (todo: remove this in post 0.9.0 release)
-	if usersCollection.OAuth2.Enabled {
-		usersCollection.OAuth2.Enabled = len(usersCollection.OAuth2.Providers) > 0
-	}
-	// allow oauth user creation if USER_CREATION is set
-	if userCreation, _ := GetEnv("USER_CREATION"); userCreation == "true" {
-		cr := "@request.context = 'oauth2'"
-		usersCollection.CreateRule = &cr
-	} else {
-		usersCollection.CreateRule = nil
-	}
-	if err := e.App.Save(usersCollection); err != nil {
-		return err
-	}
-	// allow all users to access systems if SHARE_ALL_SYSTEMS is set
-	systemsCollection, err := e.App.FindCachedCollectionByNameOrId("systems")
-	if err != nil {
-		return err
-	}
-	shareAllSystems, _ := GetEnv("SHARE_ALL_SYSTEMS")
-	systemsReadRule := "@request.auth.id != \"\""
-	if shareAllSystems != "true" {
-		// default is to only show systems that the user id is assigned to
-		systemsReadRule += " && users.id ?= @request.auth.id"
-	}
-	updateDeleteRule := systemsReadRule + " && @request.auth.role != \"readonly\""
-	systemsCollection.ListRule = &systemsReadRule
-	systemsCollection.ViewRule = &systemsReadRule
-	systemsCollection.UpdateRule = &updateDeleteRule
-	systemsCollection.DeleteRule = &updateDeleteRule
-	if err := e.App.Save(systemsCollection); err != nil {
-		return err
-	}
-	return nil
-}
-
-// startServer sets up the server for Beszel
-func (h *Hub) startServer(se *core.ServeEvent) error {
-	// TODO: exclude dev server from production binary
-	switch h.IsDev() {
-	case true:
-		proxy := httputil.NewSingleHostReverseProxy(&url.URL{
-			Scheme: "http",
-			Host:   "localhost:5173",
-		})
-		se.Router.GET("/{path...}", func(e *core.RequestEvent) error {
-			proxy.ServeHTTP(e.Response, e.Request)
-			return nil
-		})
-	default:
-		// parse app url
-		parsedURL, err := url.Parse(h.appURL)
-		if err != nil {
-			return err
-		}
-		// fix base paths in html if using subpath
-		basePath := strings.TrimSuffix(parsedURL.Path, "/") + "/"
-		indexFile, _ := fs.ReadFile(site.DistDirFS, "index.html")
-		indexContent := strings.ReplaceAll(string(indexFile), "./", basePath)
-		indexContent = strings.Replace(indexContent, "{{V}}", beszel.Version, 1)
-		indexContent = strings.Replace(indexContent, "{{HUB_URL}}", h.appURL, 1)
-		// set up static asset serving
-		staticPaths := [2]string{"/static/", "/assets/"}
-		serveStatic := apis.Static(site.DistDirFS, false)
-		// get CSP configuration
-		csp, cspExists := GetEnv("CSP")
-		// add route
-		se.Router.GET("/{path...}", func(e *core.RequestEvent) error {
-			// serve static assets if path is in staticPaths
-			for i := range staticPaths {
-				if strings.Contains(e.Request.URL.Path, staticPaths[i]) {
-					e.Response.Header().Set("Cache-Control", "public, max-age=2592000")
-					return serveStatic(e)
-				}
-			}
-			if cspExists {
-				e.Response.Header().Del("X-Frame-Options")
-				e.Response.Header().Set("Content-Security-Policy", csp)
-			}
-			return e.HTML(http.StatusOK, indexContent)
-		})
-	}
-	return nil
-}
-
-// registerCronJobs sets up scheduled tasks
-func (h *Hub) registerCronJobs(_ *core.ServeEvent) error {
-	// delete old system_stats and alerts_history records once every hour
-	h.Cron().MustAdd("delete old records", "8 * * * *", h.rm.DeleteOldRecords)
-	// create longer records every 10 minutes
-	h.Cron().MustAdd("create longer records", "*/10 * * * *", h.rm.CreateLongerRecords)
-	return nil
-}
-
-// custom api routes
-func (h *Hub) registerApiRoutes(se *core.ServeEvent) error {
-	// auth protected routes
-	apiAuth := se.Router.Group("/api/beszel")
-	apiAuth.Bind(apis.RequireAuth())
-	// auth optional routes
-	apiNoAuth := se.Router.Group("/api/beszel")
-
-	// create first user endpoint only needed if no users exist
-	if totalUsers, _ := se.App.CountRecords("users"); totalUsers == 0 {
-		apiNoAuth.POST("/create-user", h.um.CreateFirstUser)
-	}
-	// check if first time setup on login page
-	apiNoAuth.GET("/first-run", func(e *core.RequestEvent) error {
-		total, err := e.App.CountRecords("users")
-		return e.JSON(http.StatusOK, map[string]bool{"firstRun": err == nil && total == 0})
-	})
-	// get public key and version
-	apiAuth.GET("/getkey", func(e *core.RequestEvent) error {
-		return e.JSON(http.StatusOK, map[string]string{"key": h.pubKey, "v": beszel.Version})
-	})
-	// send test notification
-	apiAuth.POST("/test-notification", h.SendTestNotification)
-	// get config.yml content
-	apiAuth.GET("/config-yaml", config.GetYamlConfig)
-	// handle agent websocket connection
-	apiNoAuth.GET("/agent-connect", h.handleAgentConnect)
-	// get or create universal tokens
-	apiAuth.GET("/universal-token", h.getUniversalToken)
-	// update / delete user alerts
-	apiAuth.POST("/user-alerts", alerts.UpsertUserAlerts)
-	apiAuth.DELETE("/user-alerts", alerts.DeleteUserAlerts)
-
-	return nil
-}
-
-// Handler for universal token API endpoint (create, read, delete)
-func (h *Hub) getUniversalToken(e *core.RequestEvent) error {
-	tokenMap := universalTokenMap.GetMap()
-	userID := e.Auth.Id
-	query := e.Request.URL.Query()
-	token := query.Get("token")
-
-	if token == "" {
-		// return existing token if it exists
-		if token, _, ok := tokenMap.GetByValue(userID); ok {
-			return e.JSON(http.StatusOK, map[string]any{"token": token, "active": true})
-		}
-		// if no token is provided, generate a new one
-		token = uuid.New().String()
-	}
-	response := map[string]any{"token": token}
-
-	switch query.Get("enable") {
-	case "1":
-		tokenMap.Set(token, userID, time.Hour)
-	case "0":
-		tokenMap.RemovebyValue(userID)
-	}
-	_, response["active"] = tokenMap.GetOk(token)
-	return e.JSON(http.StatusOK, response)
-}
-
-// generates key pair if it doesn't exist and returns signer
-func (h *Hub) GetSSHKey(dataDir string) (ssh.Signer, error) {
-	if h.signer != nil {
-		return h.signer, nil
-	}
-
-	if dataDir == "" {
-		dataDir = h.DataDir()
-	}
-
-	privateKeyPath := path.Join(dataDir, "id_ed25519")
-
-	// check if the key pair already exists
-	existingKey, err := os.ReadFile(privateKeyPath)
-	if err == nil {
-		private, err := ssh.ParsePrivateKey(existingKey)
-		if err != nil {
-			return nil, fmt.Errorf("failed to parse private key: %s", err)
-		}
-		pubKeyBytes := ssh.MarshalAuthorizedKey(private.PublicKey())
-		h.pubKey = strings.TrimSuffix(string(pubKeyBytes), "\n")
-		return private, nil
-	} else if !os.IsNotExist(err) {
-		// File exists but couldn't be read for some other reason
-		return nil, fmt.Errorf("failed to read %s: %w", privateKeyPath, err)
-	}
-
-	// Generate the Ed25519 key pair
-	_, privKey, err := ed25519.GenerateKey(nil)
-	if err != nil {
-		return nil, err
-	}
-	privKeyPem, err := ssh.MarshalPrivateKey(privKey, "")
-	if err != nil {
-		return nil, err
-	}
-
-	if err := os.WriteFile(privateKeyPath, pem.EncodeToMemory(privKeyPem), 0600); err != nil {
-		return nil, fmt.Errorf("failed to write private key to %q: err: %w", privateKeyPath, err)
-	}
-
-	// These are fine to ignore the errors on, as we've literally just created a crypto.PublicKey | crypto.Signer
-	sshPrivate, _ := ssh.NewSignerFromSigner(privKey)
-	pubKeyBytes := ssh.MarshalAuthorizedKey(sshPrivate.PublicKey())
-	h.pubKey = strings.TrimSuffix(string(pubKeyBytes), "\n")
-
-	h.Logger().Info("ed25519 key pair generated successfully.")
-	h.Logger().Info("Saved to: " + privateKeyPath)
-
-	return sshPrivate, err
-}
-
-// MakeLink formats a link with the app URL and path segments.
-// Only path segments should be provided.
-func (h *Hub) MakeLink(parts ...string) string {
-	base := strings.TrimSuffix(h.Settings().Meta.AppURL, "/")
-	for _, part := range parts {
-		if part == "" {
-			continue
-		}
-		base = fmt.Sprintf("%s/%s", base, url.PathEscape(part))
-	}
-	return base
-}

beszel/internal/hub/systems/system.go

@@ -1,387 +0,0 @@
-package systems
-
-import (
-	"beszel"
-	"beszel/internal/entities/system"
-	"beszel/internal/hub/ws"
-	"context"
-	"encoding/json"
-	"errors"
-	"fmt"
-	"math/rand"
-	"net"
-	"strings"
-	"time"
-
-	"github.com/blang/semver"
-	"github.com/fxamacker/cbor/v2"
-	"github.com/pocketbase/pocketbase/core"
-	"golang.org/x/crypto/ssh"
-)
-
-type System struct {
-	Id           string               `db:"id"`
-	Host         string               `db:"host"`
-	Port         string               `db:"port"`
-	Status       string               `db:"status"`
-	manager      *SystemManager       // Manager that this system belongs to
-	client       *ssh.Client          // SSH client for fetching data
-	data         *system.CombinedData // system data from agent
-	ctx          context.Context      // Context for stopping the updater
-	cancel       context.CancelFunc   // Stops and removes system from updater
-	WsConn       *ws.WsConn           // Handler for agent WebSocket connection
-	agentVersion semver.Version       // Agent version
-	updateTicker *time.Ticker         // Ticker for updating the system
-}
-
-func (sm *SystemManager) NewSystem(systemId string) *System {
-	system := &System{
-		Id:   systemId,
-		data: &system.CombinedData{},
-	}
-	system.ctx, system.cancel = system.getContext()
-	return system
-}
-
-// StartUpdater starts the system updater.
-// It first fetches the data from the agent then updates the records.
-// If the data is not found or the system is down, it sets the system down.
-func (sys *System) StartUpdater() {
-	// Channel that can be used to set the system down. Currently only used to
-	// allow a short delay for reconnection after websocket connection is closed.
-	var downChan chan struct{}
-
-	// Add random jitter to first WebSocket connection to prevent
-	// clustering if all agents are started at the same time.
-	// SSH connections during hub startup are already staggered.
-	var jitter <-chan time.Time
-	if sys.WsConn != nil {
-		jitter = getJitter()
-		// use the websocket connection's down channel to set the system down
-		downChan = sys.WsConn.DownChan
-	} else {
-		// if the system does not have a websocket connection, wait before updating
-		// to allow the agent to connect via websocket (makes sure fingerprint is set).
-		time.Sleep(11 * time.Second)
-	}
-
-	// update immediately if system is not paused (only for ws connections)
-	// we'll wait a minute before connecting via SSH to prioritize ws connections
-	if sys.Status != paused && sys.ctx.Err() == nil {
-		if err := sys.update(); err != nil {
-			_ = sys.setDown(err)
-		}
-	}
-
-	sys.updateTicker = time.NewTicker(time.Duration(interval) * time.Millisecond)
-	// Go 1.23+ will automatically stop the ticker when the system is garbage collected, however we seem to need this or testing/synctest will block even if calling runtime.GC()
-	defer sys.updateTicker.Stop()
-
-	for {
-		select {
-		case <-sys.ctx.Done():
-			return
-		case <-sys.updateTicker.C:
-			if err := sys.update(); err != nil {
-				_ = sys.setDown(err)
-			}
-		case <-downChan:
-			sys.WsConn = nil
-			downChan = nil
-			_ = sys.setDown(nil)
-		case <-jitter:
-			sys.updateTicker.Reset(time.Duration(interval) * time.Millisecond)
-			if err := sys.update(); err != nil {
-				_ = sys.setDown(err)
-			}
-		}
-	}
-}
-
-// update updates the system data and records.
-func (sys *System) update() error {
-	if sys.Status == paused {
-		sys.handlePaused()
-		return nil
-	}
-	data, err := sys.fetchDataFromAgent()
-	if err == nil {
-		_, err = sys.createRecords(data)
-	}
-	return err
-}
-
-func (sys *System) handlePaused() {
-	if sys.WsConn == nil {
-		// if the system is paused and there's no websocket connection, remove the system
-		_ = sys.manager.RemoveSystem(sys.Id)
-	} else {
-		// Send a ping to the agent to keep the connection alive if the system is paused
-		if err := sys.WsConn.Ping(); err != nil {
-			sys.manager.hub.Logger().Warn("Failed to ping agent", "system", sys.Id, "err", err)
-			_ = sys.manager.RemoveSystem(sys.Id)
-		}
-	}
-}
-
-// createRecords updates the system record and adds system_stats and container_stats records
-func (sys *System) createRecords(data *system.CombinedData) (*core.Record, error) {
-	systemRecord, err := sys.getRecord()
-	if err != nil {
-		return nil, err
-	}
-	hub := sys.manager.hub
-	// add system_stats and container_stats records
-	systemStatsCollection, err := hub.FindCachedCollectionByNameOrId("system_stats")
-	if err != nil {
-		return nil, err
-	}
-
-	systemStatsRecord := core.NewRecord(systemStatsCollection)
-	systemStatsRecord.Set("system", systemRecord.Id)
-	systemStatsRecord.Set("stats", data.Stats)
-	systemStatsRecord.Set("type", "1m")
-	if err := hub.SaveNoValidate(systemStatsRecord); err != nil {
-		return nil, err
-	}
-	// add new container_stats record
-	if len(data.Containers) > 0 {
-		containerStatsCollection, err := hub.FindCachedCollectionByNameOrId("container_stats")
-		if err != nil {
-			return nil, err
-		}
-		containerStatsRecord := core.NewRecord(containerStatsCollection)
-		containerStatsRecord.Set("system", systemRecord.Id)
-		containerStatsRecord.Set("stats", data.Containers)
-		containerStatsRecord.Set("type", "1m")
-		if err := hub.SaveNoValidate(containerStatsRecord); err != nil {
-			return nil, err
-		}
-	}
-	// update system record (do this last because it triggers alerts and we need above records to be inserted first)
-	systemRecord.Set("status", up)
-
-	systemRecord.Set("info", data.Info)
-	if err := hub.SaveNoValidate(systemRecord); err != nil {
-		return nil, err
-	}
-	return systemRecord, nil
-}
-
-// getRecord retrieves the system record from the database.
-// If the record is not found, it removes the system from the manager.
-func (sys *System) getRecord() (*core.Record, error) {
-	record, err := sys.manager.hub.FindRecordById("systems", sys.Id)
-	if err != nil || record == nil {
-		_ = sys.manager.RemoveSystem(sys.Id)
-		return nil, err
-	}
-	return record, nil
-}
-
-// setDown marks a system as down in the database.
-// It takes the original error that caused the system to go down and returns any error
-// encountered during the process of updating the system status.
-func (sys *System) setDown(originalError error) error {
-	if sys.Status == down || sys.Status == paused {
-		return nil
-	}
-	record, err := sys.getRecord()
-	if err != nil {
-		return err
-	}
-	if originalError != nil {
-		sys.manager.hub.Logger().Error("System down", "system", record.GetString("name"), "err", originalError)
-	}
-	record.Set("status", down)
-	return sys.manager.hub.SaveNoValidate(record)
-}
-
-func (sys *System) getContext() (context.Context, context.CancelFunc) {
-	if sys.ctx == nil {
-		sys.ctx, sys.cancel = context.WithCancel(context.Background())
-	}
-	return sys.ctx, sys.cancel
-}
-
-// fetchDataFromAgent attempts to fetch data from the agent,
-// prioritizing WebSocket if available.
-func (sys *System) fetchDataFromAgent() (*system.CombinedData, error) {
-	if sys.data == nil {
-		sys.data = &system.CombinedData{}
-	}
-
-	if sys.WsConn != nil && sys.WsConn.IsConnected() {
-		wsData, err := sys.fetchDataViaWebSocket()
-		if err == nil {
-			return wsData, nil
-		}
-		// close the WebSocket connection if error and try SSH
-		sys.closeWebSocketConnection()
-	}
-
-	sshData, err := sys.fetchDataViaSSH()
-	if err != nil {
-		return nil, err
-	}
-	return sshData, nil
-}
-
-func (sys *System) fetchDataViaWebSocket() (*system.CombinedData, error) {
-	if sys.WsConn == nil || !sys.WsConn.IsConnected() {
-		return nil, errors.New("no websocket connection")
-	}
-	err := sys.WsConn.RequestSystemData(sys.data)
-	if err != nil {
-		return nil, err
-	}
-	return sys.data, nil
-}
-
-// fetchDataViaSSH handles fetching data using SSH.
-// This function encapsulates the original SSH logic.
-// It updates sys.data directly upon successful fetch.
-func (sys *System) fetchDataViaSSH() (*system.CombinedData, error) {
-	maxRetries := 1
-	for attempt := 0; attempt <= maxRetries; attempt++ {
-		if sys.client == nil || sys.Status == down {
-			if err := sys.createSSHClient(); err != nil {
-				return nil, err
-			}
-		}
-
-		session, err := sys.createSessionWithTimeout(4 * time.Second)
-		if err != nil {
-			if attempt >= maxRetries {
-				return nil, err
-			}
-			sys.manager.hub.Logger().Warn("Session closed. Retrying...", "host", sys.Host, "port", sys.Port, "err", err)
-			sys.closeSSHConnection()
-			// Reset format detection on connection failure - agent might have been upgraded
-			continue
-		}
-		defer session.Close()
-
-		stdout, err := session.StdoutPipe()
-		if err != nil {
-			return nil, err
-		}
-		if err := session.Shell(); err != nil {
-			return nil, err
-		}
-
-		*sys.data = system.CombinedData{}
-
-		if sys.agentVersion.GTE(beszel.MinVersionCbor) {
-			err = cbor.NewDecoder(stdout).Decode(sys.data)
-		} else {
-			err = json.NewDecoder(stdout).Decode(sys.data)
-		}
-
-		if err != nil {
-			sys.closeSSHConnection()
-			if attempt < maxRetries {
-				continue
-			}
-			return nil, err
-		}
-
-		// wait for the session to complete
-		if err := session.Wait(); err != nil {
-			return nil, err
-		}
-
-		return sys.data, nil
-	}
-
-	// this should never be reached due to the return in the loop
-	return nil, fmt.Errorf("failed to fetch data")
-}
-
-// createSSHClient creates a new SSH client for the system
-func (s *System) createSSHClient() error {
-	if s.manager.sshConfig == nil {
-		if err := s.manager.createSSHClientConfig(); err != nil {
-			return err
-		}
-	}
-	network := "tcp"
-	host := s.Host
-	if strings.HasPrefix(host, "/") {
-		network = "unix"
-	} else {
-		host = net.JoinHostPort(host, s.Port)
-	}
-	var err error
-	s.client, err = ssh.Dial(network, host, s.manager.sshConfig)
-	if err != nil {
-		return err
-	}
-	s.agentVersion, _ = extractAgentVersion(string(s.client.Conn.ServerVersion()))
-	return nil
-}
-
-// createSessionWithTimeout creates a new SSH session with a timeout to avoid hanging
-// in case of network issues
-func (sys *System) createSessionWithTimeout(timeout time.Duration) (*ssh.Session, error) {
-	if sys.client == nil {
-		return nil, fmt.Errorf("client not initialized")
-	}
-
-	ctx, cancel := context.WithTimeout(sys.ctx, timeout)
-	defer cancel()
-
-	sessionChan := make(chan *ssh.Session, 1)
-	errChan := make(chan error, 1)
-
-	go func() {
-		if session, err := sys.client.NewSession(); err != nil {
-			errChan <- err
-		} else {
-			sessionChan <- session
-		}
-	}()
-
-	select {
-	case session := <-sessionChan:
-		return session, nil
-	case err := <-errChan:
-		return nil, err
-	case <-ctx.Done():
-		return nil, fmt.Errorf("timeout")
-	}
-}
-
-// closeSSHConnection closes the SSH connection but keeps the system in the manager
-func (sys *System) closeSSHConnection() {
-	if sys.client != nil {
-		sys.client.Close()
-		sys.client = nil
-	}
-}
-
-// closeWebSocketConnection closes the WebSocket connection but keeps the system in the manager
-// to allow updating via SSH. It will be removed if the WS connection is re-established.
-// The system will be set as down a few seconds later if the connection is not re-established.
-func (sys *System) closeWebSocketConnection() {
-	if sys.WsConn != nil {
-		sys.WsConn.Close(nil)
-	}
-}
-
-// extractAgentVersion extracts the beszel version from SSH server version string
-func extractAgentVersion(versionString string) (semver.Version, error) {
-	_, after, _ := strings.Cut(versionString, "_")
-	return semver.Parse(after)
-}
-
-// getJitter returns a channel that will be triggered after a random delay
-// between 40% and 90% of the interval.
-// This is used to stagger the initial WebSocket connections to prevent clustering.
-func getJitter() <-chan time.Time {
-	minPercent := 40
-	maxPercent := 90
-	jitterRange := maxPercent - minPercent
-	msDelay := (interval * minPercent / 100) + rand.Intn(interval*jitterRange/100)
-	return time.After(time.Duration(msDelay) * time.Millisecond)
-}

beszel/internal/records/records_test_helpers.go

@@ -1,23 +0,0 @@
-//go:build testing
-// +build testing
-
-package records
-
-import (
-	"github.com/pocketbase/pocketbase/core"
-)
-
-// TestDeleteOldSystemStats exposes deleteOldSystemStats for testing
-func TestDeleteOldSystemStats(app core.App) error {
-	return deleteOldSystemStats(app)
-}
-
-// TestDeleteOldAlertsHistory exposes deleteOldAlertsHistory for testing
-func TestDeleteOldAlertsHistory(app core.App, countToKeep, countBeforeDeletion int) error {
-	return deleteOldAlertsHistory(app, countToKeep, countBeforeDeletion)
-}
-
-// TestTwoDecimals exposes twoDecimals for testing
-func TestTwoDecimals(value float64) float64 {
-	return twoDecimals(value)
-}

beszel/migrations/initial-settings.go

@@ -1,29 +0,0 @@
-package migrations
-
-import (
-	"github.com/pocketbase/pocketbase/core"
-	m "github.com/pocketbase/pocketbase/migrations"
-)
-
-const (
-	TempAdminEmail = "_@b.b"
-)
-
-func init() {
-	m.Register(func(app core.App) error {
-		// initial settings
-		settings := app.Settings()
-		settings.Meta.AppName = "Beszel"
-		settings.Meta.HideControls = true
-		settings.Logs.MinLevel = 4
-		if err := app.Save(settings); err != nil {
-			return err
-		}
-		// create superuser
-		collection, _ := app.FindCollectionByNameOrId(core.CollectionNameSuperusers)
-		user := core.NewRecord(collection)
-		user.SetEmail(TempAdminEmail)
-		user.SetRandomPassword()
-		return app.Save(user)
-	}, nil)
-}

beszel/site/public/static/favicon-green.svg

@@ -1 +0,0 @@
-<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 56 70" fill="#22c55e"><path d="M35 70H0V0h35q4.4 0 8.2 1.7a21.4 21.4 0 0 1 6.6 4.5q2.9 2.8 4.5 6.6Q56 16.7 56 21a15.4 15.4 0 0 1-.3 3.2 17.6 17.6 0 0 1-.2.8 19.4 19.4 0 0 1-1.5 4 17 17 0 0 1-2.4 3.4 13.5 13.5 0 0 1-2.6 2.3 12.5 12.5 0 0 1-.4.3q1.7 1 3 2.5Q53 39.1 54 41a18.3 18.3 0 0 1 1.5 4 17.4 17.4 0 0 1 .5 3 15.3 15.3 0 0 1 0 1q0 4.4-1.7 8.2a21.4 21.4 0 0 1-4.5 6.6q-2.8 2.9-6.6 4.6Q39.4 70 35 70ZM14 14v14h21a7 7 0 0 0 2.3-.3 6.6 6.6 0 0 0 .4-.2Q39 27 40 26a6.9 6.9 0 0 0 1.5-2.2q.5-1.3.5-2.8a7 7 0 0 0-.4-2.3 6.6 6.6 0 0 0-.1-.4Q40.9 17 40 16a7 7 0 0 0-2.3-1.4 6.9 6.9 0 0 0-2.5-.6 7.9 7.9 0 0 0-.2 0H14Zm0 28v14h21a7 7 0 0 0 2.3-.4 6.6 6.6 0 0 0 .4-.1Q39 54.9 40 54a7 7 0 0 0 1.5-2.2 6.9 6.9 0 0 0 .5-2.6 7.9 7.9 0 0 0 0-.2 7 7 0 0 0-.4-2.3 6.6 6.6 0 0 0-.1-.4Q40.9 45 40 44a7 7 0 0 0-2.3-1.5 6.9 6.9 0 0 0-2.5-.6 7.9 7.9 0 0 0-.2 0H14Z"/></svg>

beszel/site/public/static/favicon-red.svg

@@ -1 +0,0 @@
-<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 56 70" fill="#dc2626"><path d="M35 70H0V0h35q4.4 0 8.2 1.7a21.4 21.4 0 0 1 6.6 4.5q2.9 2.8 4.5 6.6Q56 16.7 56 21a15.4 15.4 0 0 1-.3 3.2 17.6 17.6 0 0 1-.2.8 19.4 19.4 0 0 1-1.5 4 17 17 0 0 1-2.4 3.4 13.5 13.5 0 0 1-2.6 2.3 12.5 12.5 0 0 1-.4.3q1.7 1 3 2.5Q53 39.1 54 41a18.3 18.3 0 0 1 1.5 4 17.4 17.4 0 0 1 .5 3 15.3 15.3 0 0 1 0 1q0 4.4-1.7 8.2a21.4 21.4 0 0 1-4.5 6.6q-2.8 2.9-6.6 4.6Q39.4 70 35 70ZM14 14v14h21a7 7 0 0 0 2.3-.3 6.6 6.6 0 0 0 .4-.2Q39 27 40 26a6.9 6.9 0 0 0 1.5-2.2q.5-1.3.5-2.8a7 7 0 0 0-.4-2.3 6.6 6.6 0 0 0-.1-.4Q40.9 17 40 16a7 7 0 0 0-2.3-1.4 6.9 6.9 0 0 0-2.5-.6 7.9 7.9 0 0 0-.2 0H14Zm0 28v14h21a7 7 0 0 0 2.3-.4 6.6 6.6 0 0 0 .4-.1Q39 54.9 40 54a7 7 0 0 0 1.5-2.2 6.9 6.9 0 0 0 .5-2.6 7.9 7.9 0 0 0 0-.2 7 7 0 0 0-.4-2.3 6.6 6.6 0 0 0-.1-.4Q40.9 45 40 44a7 7 0 0 0-2.3-1.5 6.9 6.9 0 0 0-2.5-.6 7.9 7.9 0 0 0-.2 0H14Z"/></svg>

beszel/site/public/static/favicon.svg

@@ -1 +0,0 @@
-<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 56 70" fill="#888"><path d="M35 70H0V0h35q4.4 0 8.2 1.7a21.4 21.4 0 0 1 6.6 4.5q2.9 2.8 4.5 6.6Q56 16.7 56 21a15.4 15.4 0 0 1-.3 3.2 17.6 17.6 0 0 1-.2.8 19.4 19.4 0 0 1-1.5 4 17 17 0 0 1-2.4 3.4 13.5 13.5 0 0 1-2.6 2.3 12.5 12.5 0 0 1-.4.3q1.7 1 3 2.5Q53 39.1 54 41a18.3 18.3 0 0 1 1.5 4 17.4 17.4 0 0 1 .5 3 15.3 15.3 0 0 1 0 1q0 4.4-1.7 8.2a21.4 21.4 0 0 1-4.5 6.6q-2.8 2.9-6.6 4.6Q39.4 70 35 70ZM14 14v14h21a7 7 0 0 0 2.3-.3 6.6 6.6 0 0 0 .4-.2Q39 27 40 26a6.9 6.9 0 0 0 1.5-2.2q.5-1.3.5-2.8a7 7 0 0 0-.4-2.3 6.6 6.6 0 0 0-.1-.4Q40.9 17 40 16a7 7 0 0 0-2.3-1.4 6.9 6.9 0 0 0-2.5-.6 7.9 7.9 0 0 0-.2 0H14Zm0 28v14h21a7 7 0 0 0 2.3-.4 6.6 6.6 0 0 0 .4-.1Q39 54.9 40 54a7 7 0 0 0 1.5-2.2 6.9 6.9 0 0 0 .5-2.6 7.9 7.9 0 0 0 0-.2 7 7 0 0 0-.4-2.3 6.6 6.6 0 0 0-.1-.4Q40.9 45 40 44a7 7 0 0 0-2.3-1.5 6.9 6.9 0 0 0-2.5-.6 7.9 7.9 0 0 0-.2 0H14Z"/></svg>

beszel/site/src/components/charts/gpu-power-chart.tsx

@@ -1,105 +0,0 @@
-import { CartesianGrid, Line, LineChart, YAxis } from "recharts"
-
-import {
-	ChartContainer,
-	ChartLegend,
-	ChartLegendContent,
-	ChartTooltip,
-	ChartTooltipContent,
-	xAxis,
-} from "@/components/ui/chart"
-import { useYAxisWidth, cn, formatShortDate, toFixedFloat, decimalString, chartMargin } from "@/lib/utils"
-import { ChartData } from "@/types"
-import { memo, useMemo } from "react"
-
-export default memo(function GpuPowerChart({ chartData }: { chartData: ChartData }) {
-	const { yAxisWidth, updateYAxisWidth } = useYAxisWidth()
-
-	if (chartData.systemStats.length === 0) {
-		return null
-	}
-
-	/** Format temperature data for chart and assign colors */
-	const newChartData = useMemo(() => {
-		const newChartData = { data: [], colors: {} } as {
-			data: Record<string, number | string>[]
-			colors: Record<string, string>
-		}
-		const powerSums = {} as Record<string, number>
-		for (let data of chartData.systemStats) {
-			let newData = { created: data.created } as Record<string, number | string>
-
-			for (let gpu of Object.values(data.stats?.g ?? {})) {
-				if (gpu.p) {
-					const name = gpu.n
-					newData[name] = gpu.p
-					powerSums[name] = (powerSums[name] ?? 0) + newData[name]
-				}
-			}
-			newChartData.data.push(newData)
-		}
-		const keys = Object.keys(powerSums).sort((a, b) => powerSums[b] - powerSums[a])
-		for (let key of keys) {
-			newChartData.colors[key] = `hsl(${((keys.indexOf(key) * 360) / keys.length) % 360}, 60%, 55%)`
-		}
-		return newChartData
-	}, [chartData])
-
-	const colors = Object.keys(newChartData.colors)
-
-	// console.log('rendered at', new Date())
-
-	return (
-		<div>
-			<ChartContainer
-				className={cn("h-full w-full absolute aspect-auto bg-card opacity-0 transition-opacity", {
-					"opacity-100": yAxisWidth,
-				})}
-			>
-				<LineChart accessibilityLayer data={newChartData.data} margin={chartMargin}>
-					<CartesianGrid vertical={false} />
-					<YAxis
-						direction="ltr"
-						orientation={chartData.orientation}
-						className="tracking-tighter"
-						domain={[0, "auto"]}
-						width={yAxisWidth}
-						tickFormatter={(value) => {
-							const val = toFixedFloat(value, 2)
-							return updateYAxisWidth(val + "W")
-						}}
-						tickLine={false}
-						axisLine={false}
-					/>
-					{xAxis(chartData)}
-					<ChartTooltip
-						animationEasing="ease-out"
-						animationDuration={150}
-						// @ts-ignore
-						itemSorter={(a, b) => b.value - a.value}
-						content={
-							<ChartTooltipContent
-								labelFormatter={(_, data) => formatShortDate(data[0].payload.created)}
-								contentFormatter={(item) => decimalString(item.value) + "W"}
-								// indicator="line"
-							/>
-						}
-					/>
-					{colors.map((key) => (
-						<Line
-							key={key}
-							dataKey={key}
-							name={key}
-							type="monotoneX"
-							dot={false}
-							strokeWidth={1.5}
-							stroke={newChartData.colors[key]}
-							isAnimationActive={false}
-						/>
-					))}
-					{colors.length > 1 && <ChartLegend content={<ChartLegendContent />} />}
-				</LineChart>
-			</ChartContainer>
-		</div>
-	)
-})

beszel/site/src/components/logo.tsx

@@ -1,17 +0,0 @@
-export function Logo({ className }: { className?: string }) {
-	return (
-		// Righteous
-		<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 285 75" className={className}>
-			{/* <defs>
-				<linearGradient id="gradient" x1="0%" y1="20%" x2="100%" y2="120%">
-					<stop offset="0%" style={{ stopColor: "#747bff" }} />
-					<stop offset="100%" style={{ stopColor: "#24eb5c" }} />
-				</linearGradient>
-			</defs> */}
-			<path
-				// fill="url(#gradient)"
-				d="M146.4 73.1h-30.5V59.8h30.5a3.2 3.2 0 0 0 2.3-1 3.2 3.2 0 0 0 1-2.3q0-.8-.3-1.3a1.5 1.5 0 0 0-.7-.6 4.7 4.7 0 0 0-1-.3l-1.3-.1h-13.9q-3.4 0-6.5-1.3-3-1.3-5.2-3.6a16.9 16.9 0 0 1-3.6-5.3 16.3 16.3 0 0 1-1.3-6.5 16.4 16.4 0 0 1 1.3-6.4q1.3-3.1 3.6-5.4 2.2-2.2 5.2-3.5a16.3 16.3 0 0 1 6.5-1.3h27v13.3h-27a3.2 3.2 0 0 0-2.3 1 3.2 3.2 0 0 0-1 2.3 3.3 3.3 0 0 0 1 2.4 3.3 3.3 0 0 0 1.2.8 3.2 3.2 0 0 0 1.1.2h13.9a18.1 18.1 0 0 1 6 1 17.3 17.3 0 0 1 .4.2q3 1.1 5.3 3.2a15.1 15.1 0 0 1 3.6 4.9 14.7 14.7 0 0 1 1.3 5.4 17.2 17.2 0 0 1 0 .9 16 16 0 0 1-1 5.8 15.4 15.4 0 0 1-.3.7 17.3 17.3 0 0 1-3.6 5.2 16.4 16.4 0 0 1-5.3 3.6 16.2 16.2 0 0 1-6.4 1.3Zm64.5-13.3v13.3h-43.6l22-39h-22V21h43.6l-22 39h22ZM35 73.1H0v-70h35q4.4 0 8.2 1.6a21.4 21.4 0 0 1 6.6 4.6q2.9 2.8 4.5 6.6 1.7 3.8 1.7 8.2a15.4 15.4 0 0 1-.3 3.2 17.6 17.6 0 0 1-.2.8 19.4 19.4 0 0 1-1.5 4 17 17 0 0 1-2.4 3.4 13.5 13.5 0 0 1-2.6 2.3 12.5 12.5 0 0 1-.4.3q1.7 1 3 2.5 1.4 1.6 2.4 3.5a18.3 18.3 0 0 1 1.5 4A17.4 17.4 0 0 1 56 51a15.3 15.3 0 0 1 0 1.1q0 4.3-1.7 8.2a21.4 21.4 0 0 1-4.5 6.6q-2.8 2.9-6.6 4.5-3.8 1.7-8.2 1.7Zm76-43L86 60.4l1.5.3a16.7 16.7 0 0 0 1.6 0q2 0 3.8-.4 1.8-.6 3.4-1.6 1.6-1 2.8-2.4a12.8 12.8 0 0 0 2-3.2l9.8 9.8q-1.9 2.6-4.3 4.7a27 27 0 0 1-5.2 3.6 26.1 26.1 0 0 1-6 2.2 26.8 26.8 0 0 1-6.3.8 26.4 26.4 0 0 1-10.4-2 26.2 26.2 0 0 1-8.5-5.8 26.7 26.7 0 0 1-5.5-8.3 30.4 30.4 0 0 1-.2-.4q-2.1-5-2.1-11.1a31.9 31.9 0 0 1 .7-7 27 27 0 0 1 1.4-4.3 27 27 0 0 1 3.8-6.6 24.5 24.5 0 0 1 2-2.2 26 26 0 0 1 8.4-5.6 27 27 0 0 1 10.4-2 26.3 26.3 0 0 1 6.4.8 26.9 26.9 0 0 1 6 2.2q2.7 1.5 5.2 3.6 2.4 2.1 4.3 4.8Zm152.3 0-25 30.2 1.5.3a16.7 16.7 0 0 0 1.6 0q2 0 3.8-.4 1.8-.6 3.4-1.6 1.5-1 2.8-2.4a12.8 12.8 0 0 0 2-3.2l9.8 9.8q-1.9 2.6-4.3 4.7a27 27 0 0 1-5.2 3.6 26.1 26.1 0 0 1-6 2.2 26.8 26.8 0 0 1-6.3.8 26.4 26.4 0 0 1-10.4-2 26.2 26.2 0 0 1-8.5-5.8A26.7 26.7 0 0 1 217 58a30.4 30.4 0 0 1-.2-.4q-2.1-5-2.1-11.1a31.9 31.9 0 0 1 .7-7 27 27 0 0 1 1.4-4.3 27 27 0 0 1 3.8-6.6 24.5 24.5 0 0 1 2-2.2 26 26 0 0 1 8.4-5.6 27 27 0 0 1 10.4-2 26.3 26.3 0 0 1 6.4.8 26.9 26.9 0 0 1 6 2.2q2.7 1.5 5.2 3.6 2.4 2.1 4.3 4.8ZM283.4 0v73.1H270V0h13.4ZM14 17v14.1h21a7 7 0 0 0 2.3-.4 6.6 6.6 0 0 0 .4-.1Q39 30 40 29a6.9 6.9 0 0 0 1.5-2.3q.5-1.3.5-2.7a7 7 0 0 0-.4-2.3 6.6 6.6 0 0 0-.1-.5q-.6-1.2-1.5-2.2a7 7 0 0 0-2.3-1.5 6.9 6.9 0 0 0-2.5-.5 7.9 7.9 0 0 0-.2 0H14Zm0 28.1v14h21a7 7 0 0 0 2.3-.4 6.6 6.6 0 0 0 .4-.2Q39 58 40 57.1a7 7 0 0 0 1.5-2.3 6.9 6.9 0 0 0 .5-2.5 7.9 7.9 0 0 0 0-.2 7 7 0 0 0-.4-2.3 6.6 6.6 0 0 0-.1-.4Q40.9 48 40 47a7 7 0 0 0-2.3-1.4 6.9 6.9 0 0 0-2.5-.6 7.9 7.9 0 0 0-.2 0H14Zm63.3 8.3 15.5-20.6a8 8 0 0 0-1.4-.4 7 7 0 0 0-.4 0 17.2 17.2 0 0 0-1.6-.1 19.2 19.2 0 0 0-.3 0 13.3 13.3 0 0 0-5.1 1q-2.5 1-4.2 2.8a13.1 13.1 0 0 0-2.5 3.6 15.5 15.5 0 0 0-.3.9 14.7 14.7 0 0 0-1 3.5 18.7 18.7 0 0 0 0 2.4 17.6 17.6 0 0 0 0 .7v.8a29.4 29.4 0 0 0 0 .1 19.2 19.2 0 0 0 .2 2 20.2 20.2 0 0 0 .4 1.6 18.6 18.6 0 0 0 0 .2 7.5 7.5 0 0 0 .4.9 6 6 0 0 0 .3.6Zm152.3 0L245 32.8a8 8 0 0 0-1.4-.4 7 7 0 0 0-.4 0 17.2 17.2 0 0 0-1.6-.1 19.2 19.2 0 0 0-.3 0 13.3 13.3 0 0 0-5.1 1q-2.5 1-4.2 2.8a13.1 13.1 0 0 0-2.5 3.6 15.5 15.5 0 0 0-.4.9 14.7 14.7 0 0 0-.8 3.5 18.7 18.7 0 0 0-.2 2.4 17.6 17.6 0 0 0 0 .7v.8a29.4 29.4 0 0 0 .1.1 19.2 19.2 0 0 0 .2 2 20.2 20.2 0 0 0 .4 1.6 18.6 18.6 0 0 0 0 .2 7.5 7.5 0 0 0 .4.9 6 6 0 0 0 .3.6Z"
-			/>
-		</svg>
-	)
-}

beszel/site/src/components/routes/home.tsx

@@ -1,141 +0,0 @@
-import { Suspense, memo, useEffect, useMemo } from "react"
-import { Card, CardContent, CardHeader, CardTitle } from "../ui/card"
-import { $alerts, $systems } from "@/lib/stores"
-import { useStore } from "@nanostores/react"
-import { GithubIcon } from "lucide-react"
-import { Separator } from "../ui/separator"
-import { getSystemNameFromId } from "@/lib/utils"
-import { pb, updateRecordList, updateSystemList } from "@/lib/api"
-import { AlertRecord, SystemRecord } from "@/types"
-import { Alert, AlertDescription, AlertTitle } from "@/components/ui/alert"
-import { $router, Link } from "../router"
-import { Plural, Trans, useLingui } from "@lingui/react/macro"
-import { getPagePath } from "@nanostores/router"
-import { alertInfo } from "@/lib/alerts"
-import SystemsTable from "@/components/systems-table/systems-table"
-
-// const SystemsTable = lazy(() => import("../systems-table/systems-table"))
-
-export default memo(function () {
-	const { t } = useLingui()
-
-	useEffect(() => {
-		document.title = t`Dashboard` + " / Beszel"
-	}, [t])
-
-	useEffect(() => {
-		// make sure we have the latest list of systems
-		updateSystemList()
-
-		// subscribe to real time updates for systems / alerts
-		pb.collection<SystemRecord>("systems").subscribe("*", (e) => {
-			updateRecordList(e, $systems)
-		})
-		return () => {
-			pb.collection("systems").unsubscribe("*")
-		}
-	}, [])
-
-	return useMemo(
-		() => (
-			<>
-				<ActiveAlerts />
-				<Suspense>
-					<SystemsTable />
-				</Suspense>
-
-				<div className="flex gap-1.5 justify-end items-center pe-3 sm:pe-6 mt-3.5 mb-4 text-xs opacity-80">
-					<a
-						href="https://github.com/henrygd/beszel"
-						target="_blank"
-						className="flex items-center gap-0.5 text-muted-foreground hover:text-foreground duration-75"
-					>
-						<GithubIcon className="h-3 w-3" /> GitHub
-					</a>
-					<Separator orientation="vertical" className="h-2.5 bg-muted-foreground opacity-70" />
-					<a
-						href="https://github.com/henrygd/beszel/releases"
-						target="_blank"
-						className="text-muted-foreground hover:text-foreground duration-75"
-					>
-						Beszel {globalThis.BESZEL.HUB_VERSION}
-					</a>
-				</div>
-			</>
-		),
-		[]
-	)
-})
-
-const ActiveAlerts = () => {
-	const alerts = useStore($alerts)
-
-	const { activeAlerts, alertsKey } = useMemo(() => {
-		const activeAlerts: AlertRecord[] = []
-		// key to prevent re-rendering if alerts change but active alerts didn't
-		const alertsKey: string[] = []
-
-		for (const systemId of Object.keys(alerts)) {
-			for (const alert of alerts[systemId].values()) {
-				if (alert.triggered && alert.name in alertInfo) {
-					activeAlerts.push(alert)
-					alertsKey.push(`${alert.system}${alert.value}${alert.min}`)
-				}
-			}
-		}
-
-		return { activeAlerts, alertsKey }
-	}, [alerts])
-
-	return useMemo(() => {
-		if (activeAlerts.length === 0) {
-			return null
-		}
-		return (
-			<Card className="mb-4">
-				<CardHeader className="pb-4 px-2 sm:px-6 max-sm:pt-5 max-sm:pb-1">
-					<div className="px-2 sm:px-1">
-						<CardTitle>
-							<Trans>Active Alerts</Trans>
-						</CardTitle>
-					</div>
-				</CardHeader>
-				<CardContent className="max-sm:p-2">
-					{activeAlerts.length > 0 && (
-						<div className="grid sm:grid-cols-2 lg:grid-cols-3 2xl:grid-cols-4 gap-3">
-							{activeAlerts.map((alert) => {
-								const info = alertInfo[alert.name as keyof typeof alertInfo]
-								return (
-									<Alert
-										key={alert.id}
-										className="hover:-translate-y-px duration-200 bg-transparent border-foreground/10 hover:shadow-md shadow-black/5"
-									>
-										<info.icon className="h-4 w-4" />
-										<AlertTitle>
-											{getSystemNameFromId(alert.system)} {info.name().toLowerCase().replace("cpu", "CPU")}
-										</AlertTitle>
-										<AlertDescription>
-											{alert.name === "Status" ? (
-												<Trans>Connection is down</Trans>
-											) : (
-												<Trans>
-													Exceeds {alert.value}
-													{info.unit} in last <Plural value={alert.min} one="# minute" other="# minutes" />
-												</Trans>
-											)}
-										</AlertDescription>
-										<Link
-											href={getPagePath($router, "system", { name: getSystemNameFromId(alert.system) })}
-											className="absolute inset-0 w-full h-full"
-											aria-label="View system"
-										></Link>
-									</Alert>
-								)
-							})}
-						</div>
-					)}
-				</CardContent>
-			</Card>
-		)
-	}, [alertsKey.join("")])
-}

beszel/site/src/components/routes/system.tsx

@@ -1,923 +0,0 @@
-import { t } from "@lingui/core/macro"
-import { Plural, Trans } from "@lingui/react/macro"
-import {
-	$systems,
-	$chartTime,
-	$containerFilter,
-	$userSettings,
-	$direction,
-	$maxValues,
-	$temperatureFilter,
-} from "@/lib/stores"
-import { ChartData, ChartTimes, ContainerStatsRecord, GPUData, SystemRecord, SystemStatsRecord } from "@/types"
-import { ChartType, Unit, Os, SystemStatus } from "@/lib/enums"
-import React, { memo, useCallback, useEffect, useMemo, useRef, useState, type JSX } from "react"
-import { Card, CardHeader, CardTitle, CardDescription } from "../ui/card"
-import { useStore } from "@nanostores/react"
-import Spinner from "../spinner"
-import { ClockArrowUp, CpuIcon, GlobeIcon, LayoutGridIcon, MonitorIcon, XIcon } from "lucide-react"
-import ChartTimeSelect from "../charts/chart-time-select"
-import {
-	chartTimeData,
-	cn,
-	decimalString,
-	formatBytes,
-	getHostDisplayValue,
-	listen,
-	parseSemVer,
-	toFixedFloat,
-	useLocalStorage,
-} from "@/lib/utils"
-import { getPbTimestamp, pb } from "@/lib/api"
-import { Separator } from "../ui/separator"
-import { Tooltip, TooltipContent, TooltipProvider, TooltipTrigger } from "../ui/tooltip"
-import { Button } from "../ui/button"
-import { Input } from "../ui/input"
-import { ChartAverage, ChartMax, Rows, TuxIcon, WindowsIcon, AppleIcon, FreeBsdIcon } from "../ui/icons"
-import { useIntersectionObserver } from "@/lib/use-intersection-observer"
-import { Select, SelectContent, SelectItem, SelectTrigger, SelectValue } from "../ui/select"
-import { timeTicks } from "d3-time"
-import { useLingui } from "@lingui/react/macro"
-import { $router, navigate } from "../router"
-import { getPagePath } from "@nanostores/router"
-import { batteryStateTranslations } from "@/lib/i18n"
-import AreaChartDefault from "@/components/charts/area-chart"
-import ContainerChart from "@/components/charts/container-chart"
-import MemChart from "@/components/charts/mem-chart"
-import DiskChart from "@/components/charts/disk-chart"
-import SwapChart from "@/components/charts/swap-chart"
-import TemperatureChart from "@/components/charts/temperature-chart"
-import GpuPowerChart from "@/components/charts/gpu-power-chart"
-import LoadAverageChart from "@/components/charts/load-average-chart"
-
-const cache = new Map<string, any>()
-
-// create ticks and domain for charts
-function getTimeData(chartTime: ChartTimes, lastCreated: number) {
-	const cached = cache.get("td")
-	if (cached && cached.chartTime === chartTime) {
-		if (!lastCreated || cached.time >= lastCreated) {
-			return cached.data
-		}
-	}
-
-	const now = new Date()
-	const startTime = chartTimeData[chartTime].getOffset(now)
-	const ticks = timeTicks(startTime, now, chartTimeData[chartTime].ticks ?? 12).map((date) => date.getTime())
-	const data = {
-		ticks,
-		domain: [chartTimeData[chartTime].getOffset(now).getTime(), now.getTime()],
-	}
-	cache.set("td", { time: now.getTime(), data, chartTime })
-	return data
-}
-
-// add empty values between records to make gaps if interval is too large
-function addEmptyValues<T extends SystemStatsRecord | ContainerStatsRecord>(
-	prevRecords: T[],
-	newRecords: T[],
-	expectedInterval: number
-) {
-	const modifiedRecords: T[] = []
-	let prevTime = (prevRecords.at(-1)?.created ?? 0) as number
-	for (let i = 0; i < newRecords.length; i++) {
-		const record = newRecords[i]
-		record.created = new Date(record.created).getTime()
-		if (prevTime) {
-			const interval = record.created - prevTime
-			// if interval is too large, add a null record
-			if (interval > expectedInterval / 2 + expectedInterval) {
-				// @ts-ignore
-				modifiedRecords.push({ created: null, stats: null })
-			}
-		}
-		prevTime = record.created
-		modifiedRecords.push(record)
-	}
-	return modifiedRecords
-}
-
-async function getStats<T>(collection: string, system: SystemRecord, chartTime: ChartTimes): Promise<T[]> {
-	const lastCached = cache.get(`${system.id}_${chartTime}_${collection}`)?.at(-1)?.created as number
-	return await pb.collection<T>(collection).getFullList({
-		filter: pb.filter("system={:id} && created > {:created} && type={:type}", {
-			id: system.id,
-			created: getPbTimestamp(chartTime, lastCached ? new Date(lastCached + 1000) : undefined),
-			type: chartTimeData[chartTime].type,
-		}),
-		fields: "created,stats",
-		sort: "created",
-	})
-}
-
-function dockerOrPodman(str: string, system: SystemRecord) {
-	if (system.info.p) {
-		str = str.replace("docker", "podman").replace("Docker", "Podman")
-	}
-	return str
-}
-
-export default function SystemDetail({ name }: { name: string }) {
-	const direction = useStore($direction)
-	const { t } = useLingui()
-	const systems = useStore($systems)
-	const chartTime = useStore($chartTime)
-	const maxValues = useStore($maxValues)
-	const [grid, setGrid] = useLocalStorage("grid", true)
-	const [system, setSystem] = useState({} as SystemRecord)
-	const [systemStats, setSystemStats] = useState([] as SystemStatsRecord[])
-	const [containerData, setContainerData] = useState([] as ChartData["containerData"])
-	const netCardRef = useRef<HTMLDivElement>(null)
-	const persistChartTime = useRef(false)
-	const [containerFilterBar, setContainerFilterBar] = useState(null as null | JSX.Element)
-	const [bottomSpacing, setBottomSpacing] = useState(0)
-	const [chartLoading, setChartLoading] = useState(true)
-	const isLongerChart = chartTime !== "1h"
-	const userSettings = $userSettings.get()
-
-	useEffect(() => {
-		document.title = `${name} / Beszel`
-		return () => {
-			if (!persistChartTime.current) {
-				$chartTime.set($userSettings.get().chartTime)
-			}
-			persistChartTime.current = false
-			setSystemStats([])
-			setContainerData([])
-			setContainerFilterBar(null)
-			$containerFilter.set("")
-		}
-	}, [name])
-
-	// function resetCharts() {
-	// 	setSystemStats([])
-	// 	setContainerData([])
-	// }
-
-	// useEffect(resetCharts, [chartTime])
-
-	// find matching system
-	useEffect(() => {
-		if (system.id && system.name === name) {
-			return
-		}
-		const matchingSystem = systems.find((s) => s.name === name) as SystemRecord
-		if (matchingSystem) {
-			setSystem(matchingSystem)
-		}
-	}, [name, system, systems])
-
-	// update system when new data is available
-	useEffect(() => {
-		if (!system.id) {
-			return
-		}
-		pb.collection<SystemRecord>("systems").subscribe(system.id, (e) => {
-			setSystem(e.record)
-		})
-		return () => {
-			pb.collection("systems").unsubscribe(system.id)
-		}
-	}, [system.id])
-
-	const chartData: ChartData = useMemo(() => {
-		const lastCreated = Math.max(
-			(systemStats.at(-1)?.created as number) ?? 0,
-			(containerData.at(-1)?.created as number) ?? 0
-		)
-		return {
-			systemStats,
-			containerData,
-			chartTime,
-			orientation: direction === "rtl" ? "right" : "left",
-			...getTimeData(chartTime, lastCreated),
-			agentVersion: parseSemVer(system?.info?.v),
-		}
-	}, [systemStats, containerData, direction])
-
-	// get stats
-	useEffect(() => {
-		if (!system.id || !chartTime) {
-			return
-		}
-		// loading: true
-		setChartLoading(true)
-		Promise.allSettled([
-			getStats<SystemStatsRecord>("system_stats", system, chartTime),
-			getStats<ContainerStatsRecord>("container_stats", system, chartTime),
-		]).then(([systemStats, containerStats]) => {
-			// loading: false
-			setChartLoading(false)
-
-			const { expectedInterval } = chartTimeData[chartTime]
-			// make new system stats
-			const ss_cache_key = `${system.id}_${chartTime}_system_stats`
-			let systemData = (cache.get(ss_cache_key) || []) as SystemStatsRecord[]
-			if (systemStats.status === "fulfilled" && systemStats.value.length) {
-				systemData = systemData.concat(addEmptyValues(systemData, systemStats.value, expectedInterval))
-				if (systemData.length > 120) {
-					systemData = systemData.slice(-100)
-				}
-				cache.set(ss_cache_key, systemData)
-			}
-			setSystemStats(systemData)
-			// make new container stats
-			const cs_cache_key = `${system.id}_${chartTime}_container_stats`
-			let containerData = (cache.get(cs_cache_key) || []) as ContainerStatsRecord[]
-			if (containerStats.status === "fulfilled" && containerStats.value.length) {
-				containerData = containerData.concat(addEmptyValues(containerData, containerStats.value, expectedInterval))
-				if (containerData.length > 120) {
-					containerData = containerData.slice(-100)
-				}
-				cache.set(cs_cache_key, containerData)
-			}
-			if (containerData.length) {
-				!containerFilterBar && setContainerFilterBar(<FilterBar />)
-			} else if (containerFilterBar) {
-				setContainerFilterBar(null)
-			}
-			makeContainerData(containerData)
-		})
-	}, [system, chartTime])
-
-	// make container stats for charts
-	const makeContainerData = useCallback((containers: ContainerStatsRecord[]) => {
-		const containerData = [] as ChartData["containerData"]
-		for (let { created, stats } of containers) {
-			if (!created) {
-				// @ts-ignore add null value for gaps
-				containerData.push({ created: null })
-				continue
-			}
-			created = new Date(created).getTime()
-			// @ts-ignore not dealing with this rn
-			let containerStats: ChartData["containerData"][0] = { created }
-			for (let container of stats) {
-				containerStats[container.n] = container
-			}
-			containerData.push(containerStats)
-		}
-		setContainerData(containerData)
-	}, [])
-
-	// values for system info bar
-	const systemInfo = useMemo(() => {
-		if (!system.info) {
-			return []
-		}
-
-		const osInfo = {
-			[Os.Linux]: {
-				Icon: TuxIcon,
-				value: system.info.k,
-				label: t({ comment: "Linux kernel", message: "Kernel" }),
-			},
-			[Os.Darwin]: {
-				Icon: AppleIcon,
-				value: `macOS ${system.info.k}`,
-			},
-			[Os.Windows]: {
-				Icon: WindowsIcon,
-				value: system.info.k,
-			},
-			[Os.FreeBSD]: {
-				Icon: FreeBsdIcon,
-				value: system.info.k,
-			},
-		}
-		let uptime: React.ReactNode
-		if (system.info.u < 3600) {
-			const mins = Math.trunc(system.info.u / 60)
-			uptime = <Plural value={mins} one="# minute" other="# minutes" />
-		} else if (system.info.u < 172800) {
-			const hours = Math.trunc(system.info.u / 3600)
-			uptime = <Plural value={hours} one="# hour" other="# hours" />
-		} else {
-			uptime = <Plural value={Math.trunc(system.info?.u / 86400)} one="# day" other="# days" />
-		}
-		return [
-			{ value: getHostDisplayValue(system), Icon: GlobeIcon },
-			{
-				value: system.info.h,
-				Icon: MonitorIcon,
-				label: "Hostname",
-				// hide if hostname is same as host or name
-				hide: system.info.h === system.host || system.info.h === system.name,
-			},
-			{ value: uptime, Icon: ClockArrowUp, label: t`Uptime`, hide: !system.info.u },
-			osInfo[system.info.os ?? Os.Linux],
-			{
-				value: `${system.info.m} (${system.info.c}c${system.info.t ? `/${system.info.t}t` : ""})`,
-				Icon: CpuIcon,
-				hide: !system.info.m,
-			},
-		] as {
-			value: string | number | undefined
-			label?: string
-			Icon: any
-			hide?: boolean
-		}[]
-	}, [system.info, t])
-
-	/** Space for tooltip if more than 12 containers */
-	useEffect(() => {
-		if (!netCardRef.current || !containerData.length) {
-			setBottomSpacing(0)
-			return
-		}
-		const tooltipHeight = (Object.keys(containerData[0]).length - 11) * 17.8 - 40
-		const wrapperEl = document.getElementById("chartwrap") as HTMLDivElement
-		const wrapperRect = wrapperEl.getBoundingClientRect()
-		const chartRect = netCardRef.current.getBoundingClientRect()
-		const distanceToBottom = wrapperRect.bottom - chartRect.bottom
-		setBottomSpacing(tooltipHeight - distanceToBottom)
-	}, [netCardRef, containerData])
-
-	// keyboard navigation between systems
-	useEffect(() => {
-		if (!systems.length) {
-			return
-		}
-		const handleKeyUp = (e: KeyboardEvent) => {
-			if (
-				e.target instanceof HTMLInputElement ||
-				e.target instanceof HTMLTextAreaElement ||
-				e.shiftKey ||
-				e.ctrlKey ||
-				e.metaKey
-			) {
-				return
-			}
-			const currentIndex = systems.findIndex((s) => s.name === name)
-			if (currentIndex === -1 || systems.length <= 1) {
-				return
-			}
-			switch (e.key) {
-				case "ArrowLeft":
-				case "h":
-					const prevIndex = (currentIndex - 1 + systems.length) % systems.length
-					persistChartTime.current = true
-					return navigate(getPagePath($router, "system", { name: systems[prevIndex].name }))
-				case "ArrowRight":
-				case "l":
-					const nextIndex = (currentIndex + 1) % systems.length
-					persistChartTime.current = true
-					return navigate(getPagePath($router, "system", { name: systems[nextIndex].name }))
-			}
-		}
-		return listen(document, "keyup", handleKeyUp)
-	}, [name, systems])
-
-	if (!system.id) {
-		return null
-	}
-
-	// select field for switching between avg and max values
-	const maxValSelect = isLongerChart ? <SelectAvgMax max={maxValues} /> : null
-	const showMax = chartTime !== "1h" && maxValues
-
-	// if no data, show empty message
-	const dataEmpty = !chartLoading && chartData.systemStats.length === 0
-	const lastGpuVals = Object.values(systemStats.at(-1)?.stats.g ?? {})
-	const hasGpuData = lastGpuVals.length > 0
-	const hasGpuPowerData = lastGpuVals.some((gpu) => gpu.p !== undefined)
-
-	let translatedStatus: string = system.status
-	if (system.status === SystemStatus.Up) {
-		translatedStatus = t({ message: "Up", comment: "Context: System is up" })
-	} else if (system.status === SystemStatus.Down) {
-		translatedStatus = t({ message: "Down", comment: "Context: System is down" })
-	}
-
-	return (
-		<>
-			<div id="chartwrap" className="grid gap-4 mb-14 overflow-x-clip">
-				{/* system info */}
-				<Card>
-					<div className="grid xl:flex gap-4 px-4 sm:px-6 pt-3 sm:pt-4 pb-5">
-						<div>
-							<h1 className="text-[1.6rem] font-semibold mb-1.5">{system.name}</h1>
-							<div className="flex flex-wrap items-center gap-3 gap-y-2 text-sm opacity-90">
-								<div className="capitalize flex gap-2 items-center">
-									<span className={cn("relative flex h-3 w-3")}>
-										{system.status === SystemStatus.Up && (
-											<span
-												className="animate-ping absolute inline-flex h-full w-full rounded-full bg-green-400 opacity-75"
-												style={{ animationDuration: "1.5s" }}
-											></span>
-										)}
-										<span
-											className={cn("relative inline-flex rounded-full h-3 w-3", {
-												"bg-green-500": system.status === SystemStatus.Up,
-												"bg-red-500": system.status === SystemStatus.Down,
-												"bg-primary/40": system.status === SystemStatus.Paused,
-												"bg-yellow-500": system.status === SystemStatus.Pending,
-											})}
-										></span>
-									</span>
-									{translatedStatus}
-								</div>
-								{systemInfo.map(({ value, label, Icon, hide }, i) => {
-									if (hide || !value) {
-										return null
-									}
-									const content = (
-										<div className="flex gap-1.5 items-center">
-											<Icon className="h-4 w-4" /> {value}
-										</div>
-									)
-									return (
-										<div key={i} className="contents">
-											<Separator orientation="vertical" className="h-4 bg-primary/30" />
-											{label ? (
-												<TooltipProvider>
-													<Tooltip delayDuration={150}>
-														<TooltipTrigger asChild>{content}</TooltipTrigger>
-														<TooltipContent>{label}</TooltipContent>
-													</Tooltip>
-												</TooltipProvider>
-											) : (
-												content
-											)}
-										</div>
-									)
-								})}
-							</div>
-						</div>
-						<div className="xl:ms-auto flex items-center gap-2 max-sm:-mb-1">
-							<ChartTimeSelect className="w-full xl:w-40" />
-							<TooltipProvider delayDuration={100}>
-								<Tooltip>
-									<TooltipTrigger asChild>
-										<Button
-											aria-label={t`Toggle grid`}
-											variant="outline"
-											size="icon"
-											className="hidden xl:flex p-0 text-primary"
-											onClick={() => setGrid(!grid)}
-										>
-											{grid ? (
-												<LayoutGridIcon className="h-[1.2rem] w-[1.2rem] opacity-75" />
-											) : (
-												<Rows className="h-[1.3rem] w-[1.3rem] opacity-75" />
-											)}
-										</Button>
-									</TooltipTrigger>
-									<TooltipContent>{t`Toggle grid`}</TooltipContent>
-								</Tooltip>
-							</TooltipProvider>
-						</div>
-					</div>
-				</Card>
-
-				{/* main charts */}
-				<div className="grid xl:grid-cols-2 gap-4">
-					<ChartCard
-						empty={dataEmpty}
-						grid={grid}
-						title={t`CPU Usage`}
-						description={t`Average system-wide CPU utilization`}
-						cornerEl={maxValSelect}
-					>
-						<AreaChartDefault
-							chartData={chartData}
-							maxToggled={maxValues}
-							dataPoints={[
-								{
-									label: t`CPU Usage`,
-									dataKey: ({ stats }) => (showMax ? stats?.cpum : stats?.cpu),
-									color: 1,
-									opacity: 0.4,
-								},
-							]}
-							tickFormatter={(val) => toFixedFloat(val, 2) + "%"}
-							contentFormatter={({ value }) => decimalString(value) + "%"}
-						/>
-					</ChartCard>
-
-					{containerFilterBar && (
-						<ChartCard
-							empty={dataEmpty}
-							grid={grid}
-							title={dockerOrPodman(t`Docker CPU Usage`, system)}
-							description={t`Average CPU utilization of containers`}
-							cornerEl={containerFilterBar}
-						>
-							<ContainerChart chartData={chartData} dataKey="c" chartType={ChartType.CPU} />
-						</ChartCard>
-					)}
-
-					<ChartCard
-						empty={dataEmpty}
-						grid={grid}
-						title={t`Memory Usage`}
-						description={t`Precise utilization at the recorded time`}
-						cornerEl={maxValSelect}
-					>
-						<MemChart chartData={chartData} showMax={showMax} />
-					</ChartCard>
-
-					{containerFilterBar && (
-						<ChartCard
-							empty={dataEmpty}
-							grid={grid}
-							title={dockerOrPodman(t`Docker Memory Usage`, system)}
-							description={dockerOrPodman(t`Memory usage of docker containers`, system)}
-							cornerEl={containerFilterBar}
-						>
-							<ContainerChart chartData={chartData} dataKey="m" chartType={ChartType.Memory} />
-						</ChartCard>
-					)}
-
-					<ChartCard empty={dataEmpty} grid={grid} title={t`Disk Usage`} description={t`Usage of root partition`}>
-						<DiskChart chartData={chartData} dataKey="stats.du" diskSize={systemStats.at(-1)?.stats.d ?? NaN} />
-					</ChartCard>
-
-					<ChartCard
-						empty={dataEmpty}
-						grid={grid}
-						title={t`Disk I/O`}
-						description={t`Throughput of root filesystem`}
-						cornerEl={maxValSelect}
-					>
-						<AreaChartDefault
-							chartData={chartData}
-							maxToggled={maxValues}
-							dataPoints={[
-								{
-									label: t({ message: "Write", comment: "Disk write" }),
-									dataKey: ({ stats }) => (showMax ? stats?.dwm : stats?.dw),
-									color: 3,
-									opacity: 0.3,
-								},
-								{
-									label: t({ message: "Read", comment: "Disk read" }),
-									dataKey: ({ stats }) => (showMax ? stats?.drm : stats?.dr),
-									color: 1,
-									opacity: 0.3,
-								},
-							]}
-							tickFormatter={(val) => {
-								const { value, unit } = formatBytes(val, true, userSettings.unitDisk, true)
-								return toFixedFloat(value, value >= 10 ? 0 : 1) + " " + unit
-							}}
-							contentFormatter={({ value }) => {
-								const { value: convertedValue, unit } = formatBytes(value, true, userSettings.unitDisk, true)
-								return decimalString(convertedValue, convertedValue >= 100 ? 1 : 2) + " " + unit
-							}}
-						/>
-					</ChartCard>
-
-					<ChartCard
-						empty={dataEmpty}
-						grid={grid}
-						title={t`Bandwidth`}
-						cornerEl={maxValSelect}
-						description={t`Network traffic of public interfaces`}
-					>
-						<AreaChartDefault
-							chartData={chartData}
-							maxToggled={maxValues}
-							dataPoints={[
-								{
-									label: t`Sent`,
-									// use bytes if available, otherwise multiply old MB (can remove in future)
-									dataKey(data) {
-										if (showMax) {
-											return data?.stats?.bm?.[0] ?? (data?.stats?.nsm ?? 0) * 1024 * 1024
-										}
-										return data?.stats?.b?.[0] ?? data?.stats?.ns * 1024 * 1024
-									},
-									color: 5,
-									opacity: 0.2,
-								},
-								{
-									label: t`Received`,
-									dataKey(data) {
-										if (showMax) {
-											return data?.stats?.bm?.[1] ?? (data?.stats?.nrm ?? 0) * 1024 * 1024
-										}
-										return data?.stats?.b?.[1] ?? data?.stats?.nr * 1024 * 1024
-									},
-									color: 2,
-									opacity: 0.2,
-								},
-							]}
-							tickFormatter={(val) => {
-								let { value, unit } = formatBytes(val, true, userSettings.unitNet, false)
-								return toFixedFloat(value, value >= 10 ? 0 : 1) + " " + unit
-							}}
-							contentFormatter={(data) => {
-								const { value, unit } = formatBytes(data.value, true, userSettings.unitNet, false)
-								return decimalString(value, value >= 100 ? 1 : 2) + " " + unit
-							}}
-						/>
-					</ChartCard>
-
-					{containerFilterBar && containerData.length > 0 && (
-						<div
-							ref={netCardRef}
-							className={cn({
-								"col-span-full": !grid,
-							})}
-						>
-							<ChartCard
-								empty={dataEmpty}
-								title={dockerOrPodman(t`Docker Network I/O`, system)}
-								description={dockerOrPodman(t`Network traffic of docker containers`, system)}
-								cornerEl={containerFilterBar}
-							>
-								{/* @ts-ignore */}
-								<ContainerChart chartData={chartData} chartType={ChartType.Network} dataKey="n" />
-							</ChartCard>
-						</div>
-					)}
-
-					{/* Swap chart */}
-					{(systemStats.at(-1)?.stats.su ?? 0) > 0 && (
-						<ChartCard
-							empty={dataEmpty}
-							grid={grid}
-							title={t`Swap Usage`}
-							description={t`Swap space used by the system`}
-						>
-							<SwapChart chartData={chartData} />
-						</ChartCard>
-					)}
-
-					{/* Load Average chart */}
-					{chartData.agentVersion?.minor >= 12 && (
-						<ChartCard
-							empty={dataEmpty}
-							grid={grid}
-							title={t`Load Average`}
-							description={t`System load averages over time`}
-						>
-							<LoadAverageChart chartData={chartData} />
-						</ChartCard>
-					)}
-
-					{/* Temperature chart */}
-					{systemStats.at(-1)?.stats.t && (
-						<ChartCard
-							empty={dataEmpty}
-							grid={grid}
-							title={t`Temperature`}
-							description={t`Temperatures of system sensors`}
-							cornerEl={<FilterBar store={$temperatureFilter} />}
-						>
-							<TemperatureChart chartData={chartData} />
-						</ChartCard>
-					)}
-
-					{/* Battery chart */}
-					{systemStats.at(-1)?.stats.bat && (
-						<ChartCard
-							empty={dataEmpty}
-							grid={grid}
-							title={t`Battery`}
-							description={`${t({
-								message: "Current state",
-								comment: "Context: Battery state",
-							})}: ${batteryStateTranslations[systemStats.at(-1)?.stats.bat![1] ?? 0]()}`}
-						>
-							<AreaChartDefault
-								chartData={chartData}
-								maxToggled={maxValues}
-								dataPoints={[
-									{
-										label: t`Charge`,
-										dataKey: ({ stats }) => stats?.bat?.[0],
-										color: 1,
-										opacity: 0.35,
-									},
-								]}
-								domain={[0, 100]}
-								tickFormatter={(val) => `${val}%`}
-								contentFormatter={({ value }) => `${value}%`}
-							/>
-						</ChartCard>
-					)}
-
-					{/* GPU power draw chart */}
-					{hasGpuPowerData && (
-						<ChartCard
-							empty={dataEmpty}
-							grid={grid}
-							title={t`GPU Power Draw`}
-							description={t`Average power consumption of GPUs`}
-						>
-							<GpuPowerChart chartData={chartData} />
-						</ChartCard>
-					)}
-				</div>
-
-				{/* GPU charts */}
-				{hasGpuData && (
-					<div className="grid xl:grid-cols-2 gap-4">
-						{Object.keys(systemStats.at(-1)?.stats.g ?? {}).map((id) => {
-							const gpu = systemStats.at(-1)?.stats.g?.[id] as GPUData
-							return (
-								<div key={id} className="contents">
-									<ChartCard
-										empty={dataEmpty}
-										grid={grid}
-										title={`${gpu.n} ${t`Usage`}`}
-										description={t`Average utilization of ${gpu.n}`}
-									>
-										<AreaChartDefault
-											chartData={chartData}
-											dataPoints={[
-												{
-													label: t`Usage`,
-													dataKey: ({ stats }) => stats?.g?.[id]?.u ?? 0,
-													color: 1,
-													opacity: 0.35,
-												},
-											]}
-											tickFormatter={(val) => toFixedFloat(val, 2) + "%"}
-											contentFormatter={({ value }) => decimalString(value) + "%"}
-										/>
-									</ChartCard>
-									<ChartCard
-										empty={dataEmpty}
-										grid={grid}
-										title={`${gpu.n} VRAM`}
-										description={t`Precise utilization at the recorded time`}
-									>
-										<AreaChartDefault
-											chartData={chartData}
-											dataPoints={[
-												{
-													label: t`Usage`,
-													dataKey: ({ stats }) => stats?.g?.[id]?.mu ?? 0,
-													color: 2,
-													opacity: 0.25,
-												},
-											]}
-											max={gpu.mt}
-											tickFormatter={(val) => {
-												const { value, unit } = formatBytes(val, false, Unit.Bytes, true)
-												return toFixedFloat(value, value >= 10 ? 0 : 1) + " " + unit
-											}}
-											contentFormatter={({ value }) => {
-												const { value: convertedValue, unit } = formatBytes(value, false, Unit.Bytes, true)
-												return decimalString(convertedValue) + " " + unit
-											}}
-										/>
-									</ChartCard>
-								</div>
-							)
-						})}
-					</div>
-				)}
-
-				{/* extra filesystem charts */}
-				{Object.keys(systemStats.at(-1)?.stats.efs ?? {}).length > 0 && (
-					<div className="grid xl:grid-cols-2 gap-4">
-						{Object.keys(systemStats.at(-1)?.stats.efs ?? {}).map((extraFsName) => {
-							return (
-								<div key={extraFsName} className="contents">
-									<ChartCard
-										empty={dataEmpty}
-										grid={grid}
-										title={`${extraFsName} ${t`Usage`}`}
-										description={t`Disk usage of ${extraFsName}`}
-									>
-										<DiskChart
-											chartData={chartData}
-											dataKey={`stats.efs.${extraFsName}.du`}
-											diskSize={systemStats.at(-1)?.stats.efs?.[extraFsName].d ?? NaN}
-										/>
-									</ChartCard>
-									<ChartCard
-										empty={dataEmpty}
-										grid={grid}
-										title={`${extraFsName} I/O`}
-										description={t`Throughput of ${extraFsName}`}
-										cornerEl={maxValSelect}
-									>
-										<AreaChartDefault
-											chartData={chartData}
-											dataPoints={[
-												{
-													label: t`Write`,
-													dataKey: ({ stats }) => stats?.efs?.[extraFsName]?.[showMax ? "wm" : "w"] ?? 0,
-													color: 3,
-													opacity: 0.3,
-												},
-												{
-													label: t`Read`,
-													dataKey: ({ stats }) => stats?.efs?.[extraFsName]?.[showMax ? "rm" : "r"] ?? 0,
-													color: 1,
-													opacity: 0.3,
-												},
-											]}
-											maxToggled={maxValues}
-											tickFormatter={(val) => {
-												const { value, unit } = formatBytes(val, true, userSettings.unitDisk, true)
-												return toFixedFloat(value, value >= 10 ? 0 : 1) + " " + unit
-											}}
-											contentFormatter={({ value }) => {
-												const { value: convertedValue, unit } = formatBytes(value, true, userSettings.unitDisk, true)
-												return decimalString(convertedValue, convertedValue >= 100 ? 1 : 2) + " " + unit
-											}}
-										/>
-									</ChartCard>
-								</div>
-							)
-						})}
-					</div>
-				)}
-			</div>
-
-			{/* add space for tooltip if more than 12 containers */}
-			{bottomSpacing > 0 && <span className="block" style={{ height: bottomSpacing }} />}
-		</>
-	)
-}
-
-function FilterBar({ store = $containerFilter }: { store?: typeof $containerFilter }) {
-	const containerFilter = useStore(store)
-	const { t } = useLingui()
-
-	const handleChange = useCallback((e: React.ChangeEvent<HTMLInputElement>) => {
-		store.set(e.target.value)
-	}, [])
-
-	return (
-		<>
-			<Input placeholder={t`Filter...`} className="ps-4 pe-8" value={containerFilter} onChange={handleChange} />
-			{containerFilter && (
-				<Button
-					type="button"
-					variant="ghost"
-					size="icon"
-					aria-label="Clear"
-					className="absolute right-1 top-1/2 -translate-y-1/2 h-7 w-7 text-gray-500 hover:text-gray-900 dark:text-gray-400 dark:hover:text-gray-100"
-					onClick={() => store.set("")}
-				>
-					<XIcon className="h-4 w-4" />
-				</Button>
-			)}
-		</>
-	)
-}
-
-const SelectAvgMax = memo(({ max }: { max: boolean }) => {
-	const Icon = max ? ChartMax : ChartAverage
-	return (
-		<Select value={max ? "max" : "avg"} onValueChange={(e) => $maxValues.set(e === "max")}>
-			<SelectTrigger className="relative ps-10 pe-5">
-				<Icon className="h-4 w-4 absolute start-4 top-1/2 -translate-y-1/2 opacity-85" />
-				<SelectValue />
-			</SelectTrigger>
-			<SelectContent>
-				<SelectItem key="avg" value="avg">
-					<Trans>Average</Trans>
-				</SelectItem>
-				<SelectItem key="max" value="max">
-					<Trans comment="Chart select field. Please try to keep this short.">Max 1 min</Trans>
-				</SelectItem>
-			</SelectContent>
-		</Select>
-	)
-})
-
-function ChartCard({
-	title,
-	description,
-	children,
-	grid,
-	empty,
-	cornerEl,
-}: {
-	title: string
-	description: string
-	children: React.ReactNode
-	grid?: boolean
-	empty?: boolean
-	cornerEl?: JSX.Element | null
-}) {
-	const { isIntersecting, ref } = useIntersectionObserver()
-
-	return (
-		<Card className={cn("pb-2 sm:pb-4 odd:last-of-type:col-span-full", { "col-span-full": !grid })} ref={ref}>
-			<CardHeader className="pb-5 pt-4 gap-1 relative max-sm:py-3 max-sm:px-4">
-				<CardTitle className="text-xl sm:text-2xl">{title}</CardTitle>
-				<CardDescription>{description}</CardDescription>
-				{cornerEl && <div className="relative py-1 block sm:w-44 sm:absolute sm:top-3.5 sm:end-3.5">{cornerEl}</div>}
-			</CardHeader>
-			<div className="ps-0 w-[calc(100%-1.5em)] h-48 md:h-52 relative group">
-				{
-					<Spinner
-						msg={empty ? t`Waiting for enough records to display` : undefined}
-						// className="group-has-[.opacity-100]:opacity-0 transition-opacity"
-						className="group-has-[.opacity-100]:invisible duration-100"
-					/>
-				}
-				{isIntersecting && children}
-			</div>
-		</Card>
-	)
-}

beszel/site/src/components/ui/toaster.tsx

@@ -1,24 +0,0 @@
-import { Toast, ToastClose, ToastDescription, ToastProvider, ToastTitle, ToastViewport } from "@/components/ui/toast"
-import { useToast } from "@/components/ui/use-toast"
-
-export function Toaster() {
-	const { toasts } = useToast()
-
-	return (
-		<ToastProvider>
-			{toasts.map(function ({ id, title, description, action, ...props }) {
-				return (
-					<Toast key={id} {...props}>
-						<div className="grid gap-1">
-							{title && <ToastTitle>{title}</ToastTitle>}
-							{description && <ToastDescription>{description}</ToastDescription>}
-						</div>
-						{action}
-						<ToastClose />
-					</Toast>
-				)
-			})}
-			<ToastViewport />
-		</ToastProvider>
-	)
-}

beszel/site/src/lib/api.ts

@@ -1,136 +0,0 @@
-import { ChartTimes, SystemRecord, UserSettings } from "@/types"
-import { $alerts, $longestSystemNameLen, $systems, $userSettings } from "./stores"
-import { toast } from "@/components/ui/use-toast"
-import { t } from "@lingui/core/macro"
-import { chartTimeData } from "./utils"
-import { WritableAtom } from "nanostores"
-import { RecordModel, RecordSubscription } from "pocketbase"
-import PocketBase from "pocketbase"
-import { basePath } from "@/components/router"
-
-/** PocketBase JS Client */
-export const pb = new PocketBase(basePath)
-
-export const isAdmin = () => pb.authStore.record?.role === "admin"
-export const isReadOnlyUser = () => pb.authStore.record?.role === "readonly"
-
-const verifyAuth = () => {
-	pb.collection("users")
-		.authRefresh()
-		.catch(() => {
-			logOut()
-			toast({
-				title: t`Failed to authenticate`,
-				description: t`Please log in again`,
-				variant: "destructive",
-			})
-		})
-}
-
-/** Logs the user out by clearing the auth store and unsubscribing from realtime updates. */
-export async function logOut() {
-	$systems.set([])
-	$alerts.set({})
-	$userSettings.set({} as UserSettings)
-	sessionStorage.setItem("lo", "t") // prevent auto login on logout
-	pb.authStore.clear()
-	pb.realtime.unsubscribe()
-}
-
-/** Fetch or create user settings in database */
-export async function updateUserSettings() {
-	try {
-		const req = await pb.collection("user_settings").getFirstListItem("", { fields: "settings" })
-		$userSettings.set(req.settings)
-		return
-	} catch (e) {
-		console.error("get settings", e)
-	}
-	// create user settings if error fetching existing
-	try {
-		const createdSettings = await pb.collection("user_settings").create({ user: pb.authStore.record!.id })
-		$userSettings.set(createdSettings.settings)
-	} catch (e) {
-		console.error("create settings", e)
-	}
-}
-/** Update systems / alerts list when records change  */
-export function updateRecordList<T extends RecordModel>(e: RecordSubscription<T>, $store: WritableAtom<T[]>) {
-	const curRecords = $store.get()
-	const newRecords = []
-	if (e.action === "delete") {
-		for (const server of curRecords) {
-			if (server.id !== e.record.id) {
-				newRecords.push(server)
-			}
-		}
-	} else {
-		let found = 0
-		for (const server of curRecords) {
-			if (server.id === e.record.id) {
-				found = newRecords.push(e.record)
-			} else {
-				newRecords.push(server)
-			}
-		}
-		if (!found) {
-			newRecords.push(e.record)
-		}
-	}
-	$store.set(newRecords)
-}
-/** Fetches updated system list from database */
-export const updateSystemList = (() => {
-	let isFetchingSystems = false
-	return async () => {
-		if (isFetchingSystems) {
-			return
-		}
-		isFetchingSystems = true
-		try {
-			let records = await pb
-				.collection<SystemRecord>("systems")
-				.getFullList({ sort: "+name", fields: "id,name,host,port,info,status" })
-
-			if (records.length) {
-				// records = [
-				// 	...records,
-				// 	...records,
-				// 	...records,
-				// 	...records,
-				// 	...records,
-				// 	...records,
-				// 	...records,
-				// 	...records,
-				// 	...records,
-				// ]
-				// we need to loop once to get the longest name
-				let longestName = $longestSystemNameLen.get()
-				for (const { name } of records) {
-					const nameLen = Math.min(20, name.length)
-					if (nameLen > longestName) {
-						$longestSystemNameLen.set(nameLen)
-						longestName = nameLen
-					}
-				}
-				$systems.set(records)
-			} else {
-				verifyAuth()
-			}
-		} finally {
-			isFetchingSystems = false
-		}
-	}
-})()
-
-export function getPbTimestamp(timeString: ChartTimes, d?: Date) {
-	d ||= chartTimeData[timeString].getOffset(new Date())
-	const year = d.getUTCFullYear()
-	const month = String(d.getUTCMonth() + 1).padStart(2, "0")
-	const day = String(d.getUTCDate()).padStart(2, "0")
-	const hours = String(d.getUTCHours()).padStart(2, "0")
-	const minutes = String(d.getUTCMinutes()).padStart(2, "0")
-	const seconds = String(d.getUTCSeconds()).padStart(2, "0")
-
-	return `${year}-${month}-${day} ${hours}:${minutes}:${seconds}`
-}

beszel/site/src/lib/enums.ts

@@ -1,48 +0,0 @@
-/** Operating system */
-export enum Os {
-	Linux = 0,
-	Darwin,
-	Windows,
-	FreeBSD,
-}
-
-/** Type of chart */
-export enum ChartType {
-	Memory,
-	Disk,
-	Network,
-	CPU,
-}
-
-/** Unit of measurement */
-export enum Unit {
-	Bytes,
-	Bits,
-	Celsius,
-	Fahrenheit,
-}
-
-/** Meter state for color */
-export enum MeterState {
-	Good,
-	Warn,
-	Crit,
-}
-
-/** System status states */
-export enum SystemStatus {
-	Up = "up",
-	Down = "down",
-	Pending = "pending",
-	Paused = "paused",
-}
-
-/** Battery state */
-export enum BatteryState {
-	Unknown,
-	Empty,
-	Full,
-	Charging,
-	Discharging,
-	Idle,
-}

beszel/site/src/lib/languages.ts

@@ -1,137 +0,0 @@
-export default [
-	{
-		lang: "ar",
-		label: "العربية",
-		e: "🇵🇸",
-	},
-	{
-		lang: "bg",
-		label: "Български",
-		e: "🇧🇬",
-	},
-	{
-		lang: "cs",
-		label: "Čeština",
-		e: "🇨🇿",
-	},
-	{
-		lang: "da",
-		label: "Dansk",
-		e: "🇩🇰",
-	},
-	{
-		lang: "de",
-		label: "Deutsch",
-		e: "🇩🇪",
-	},
-	{
-		lang: "en",
-		label: "English",
-		e: "🇺🇸",
-	},
-	{
-		lang: "es",
-		label: "Español",
-		e: "🇲🇽",
-	},
-	{
-		lang: "fa",
-		label: "فارسی",
-		e: "🇮🇷",
-	},
-	{
-		lang: "fr",
-		label: "Français",
-		e: "🇫🇷",
-	},
-	{
-		lang: "hr",
-		label: "Hrvatski",
-		e: "🇭🇷",
-	},
-	{
-		lang: "hu",
-		label: "Magyar",
-		e: "🇭🇺",
-	},
-	{
-		lang: "it",
-		label: "Italiano",
-		e: "🇮🇹",
-	},
-	{
-		lang: "ja",
-		label: "日本語",
-		e: "🇯🇵",
-	},
-	{
-		lang: "ko",
-		label: "한국어",
-		e: "🇰🇷",
-	},
-	{
-		lang: "nl",
-		label: "Nederlands",
-		e: "🇳🇱",
-	},
-	{
-		lang: "no",
-		label: "Norsk",
-		e: "🇳🇴",
-	},
-	{
-		lang: "pl",
-		label: "Polski",
-		e: "🇵🇱",
-	},
-	{
-		lang: "pt",
-		label: "Português",
-		e: "🇧🇷",
-	},
-	{
-		lang: "tr",
-		label: "Türkçe",
-		e: "🇹🇷",
-	},
-	{
-		lang: "ru",
-		label: "Русский",
-		e: "🇷🇺",
-	},
-	{
-		lang: "sl",
-		label: "Slovenščina",
-		e: "🇸🇮",
-	},
-	{
-		lang: "sv",
-		label: "Svenska",
-		e: "🇸🇪",
-	},
-	{
-		lang: "uk",
-		label: "Українська",
-		e: "🇺🇦",
-	},
-	{
-		lang: "vi",
-		label: "Tiếng Việt",
-		e: "🇻🇳",
-	},
-	{
-		lang: "zh-CN",
-		label: "简体中文",
-		e: "🇨🇳",
-	},
-	{
-		lang: "zh-HK",
-		label: "繁體中文",
-		e: "🇭🇰",
-	},
-	{
-		lang: "zh",
-		label: "繁體中文",
-		e: "🇹🇼",
-	},
-] as const