lee/puter
1
0
Fork 0
mirror of https://github.com/HeyPuter/puter.git synced 2026-05-04 08:30:39 +00:00
Code Issues Actions 8 Packages Projects Releases Wiki Activity

[PUT-471] feat: estimate and bill input for gemini imgaes :dev: (#2453)

Browse Source 
* [PUT-471] feat: estimate and bill input for gemini imgaes :dev:

* [PUT-472] add input/output costs to openai image models too

* npm version bump

* remove second url log

* fix gemini pro image

---------

Co-authored-by: ProgrammerIn-wonderland <3838shah@gmail.com>
This commit is contained in:
Daniel Salazar
2026-02-10 12:52:33 -08:00
committed by GitHub
parent febcceb042
commit 91afa2c356
10 changed files with 496 additions and 107 deletions
Download Patch File Download Diff File Expand all files Collapse all files
package-lock.json
Generated
+3 -30
View File
@@ -406,6 +406,7 @@
"resolved": "https://registry.npmjs.org/@aws-sdk/client-dynamodb/-/client-dynamodb-3.980.0.tgz",
"integrity": "sha512-1rGhAx4cHZy3pMB3R3r84qMT5WEvQ6ajr2UksnD48fjQxwaUcpI6NsPvU5j/5BI5LqGiUO6ThOrMwSMm95twQA==",
"license": "Apache-2.0",
"peer": true,
"dependencies": {
"@aws-crypto/sha256-browser": "5.2.0",
"@aws-crypto/sha256-js": "5.2.0",
@@ -459,6 +460,7 @@
"resolved": "https://registry.npmjs.org/@aws-sdk/util-endpoints/-/util-endpoints-3.980.0.tgz",
"integrity": "sha512-AjKBNEc+rjOZQE1HwcD9aCELqg1GmUj1rtICKuY8cgwB73xJ4U/kNyqKKpN2k9emGqlfDY2D8itIp/vDc6OKpw==",
"license": "Apache-2.0",
"peer": true,
"dependencies": {
"@aws-sdk/types": "^3.973.1",
"@smithy/types": "^4.12.0",
@@ -1303,7 +1305,6 @@
"version": "7.28.5",
"dev": true,
"license": "MIT",
"peer": true,
"dependencies": {
"@babel/code-frame": "^7.27.1",
"@babel/generator": "^7.28.5",
@@ -1620,7 +1621,6 @@
}
],
"license": "MIT",
"peer": true,
"engines": {
"node": ">=18"
},
@@ -1659,7 +1659,6 @@
}
],
"license": "MIT",
"peer": true,
"engines": {
"node": ">=18"
}
@@ -3432,7 +3431,6 @@
"node_modules/@jimp/custom": {
"version": "0.22.12",
"license": "MIT",
"peer": true,
"dependencies": {
"@jimp/core": "^0.22.12"
}
@@ -3463,7 +3461,6 @@
"node_modules/@jimp/plugin-blit": {
"version": "0.22.12",
"license": "MIT",
"peer": true,
"dependencies": {
"@jimp/utils": "^0.22.12"
},
@@ -3474,7 +3471,6 @@
"node_modules/@jimp/plugin-blur": {
"version": "0.22.12",
"license": "MIT",
"peer": true,
"dependencies": {
"@jimp/utils": "^0.22.12"
},
@@ -3495,7 +3491,6 @@
"node_modules/@jimp/plugin-color": {
"version": "0.22.12",
"license": "MIT",
"peer": true,
"dependencies": {
"@jimp/utils": "^0.22.12",
"tinycolor2": "^1.6.0"
@@ -3533,7 +3528,6 @@
"node_modules/@jimp/plugin-crop": {
"version": "0.22.12",
"license": "MIT",
"peer": true,
"dependencies": {
"@jimp/utils": "^0.22.12"
},
@@ -3637,7 +3631,6 @@
"node_modules/@jimp/plugin-resize": {
"version": "0.22.12",
"license": "MIT",
"peer": true,
"dependencies": {
"@jimp/utils": "^0.22.12"
},
@@ -3648,7 +3641,6 @@
"node_modules/@jimp/plugin-rotate": {
"version": "0.22.12",
"license": "MIT",
"peer": true,
"dependencies": {
"@jimp/utils": "^0.22.12"
},
@@ -3662,7 +3654,6 @@
"node_modules/@jimp/plugin-scale": {
"version": "0.22.12",
"license": "MIT",
"peer": true,
"dependencies": {
"@jimp/utils": "^0.22.12"
},
@@ -3903,7 +3894,6 @@
"node_modules/@opentelemetry/api": {
"version": "1.9.0",
"license": "Apache-2.0",
"peer": true,
"engines": {
"node": ">=8.0.0"
}
@@ -5736,7 +5726,6 @@
"version": "8.48.0",
"dev": true,
"license": "MIT",
"peer": true,
"dependencies": {
"@typescript-eslint/scope-manager": "8.48.0",
"@typescript-eslint/types": "8.48.0",
@@ -6368,7 +6357,6 @@
"node_modules/acorn": {
"version": "8.15.0",
"license": "MIT",
"peer": true,
"bin": {
"acorn": "bin/acorn"
},
@@ -7026,7 +7014,6 @@
}
],
"license": "MIT",
"peer": true,
"dependencies": {
"baseline-browser-mapping": "^2.8.25",
"caniuse-lite": "^1.0.30001754",
@@ -7225,7 +7212,6 @@
"node_modules/chai": {
"version": "4.5.0",
"license": "MIT",
"peer": true,
"dependencies": {
"assertion-error": "^1.1.0",
"check-error": "^1.0.3",
@@ -8519,7 +8505,6 @@
"node_modules/eslint": {
"version": "9.39.1",
"license": "MIT",
"peer": true,
"dependencies": {
"@eslint-community/eslint-utils": "^4.8.0",
"@eslint-community/regexpp": "^4.12.1",
@@ -10478,7 +10463,6 @@
"resolved": "https://registry.npmjs.org/ioredis/-/ioredis-5.9.2.tgz",
"integrity": "sha512-tAAg/72/VxOUW7RQSX1pIxJVucYKcjFjfvj60L57jrZpYCHC3XN0WCQ3sNYL4Gmvv+7GPvTAjc+KSdeNuE8oWQ==",
"license": "MIT",
"peer": true,
"dependencies": {
"@ioredis/commands": "1.5.0",
"cluster-key-slot": "^1.1.0",
@@ -14341,7 +14325,6 @@
"version": "8.17.1",
"dev": true,
"license": "MIT",
"peer": true,
"dependencies": {
"fast-deep-equal": "^3.1.3",
"fast-uri": "^3.0.1",
@@ -15783,7 +15766,6 @@
"version": "5.9.3",
"dev": true,
"license": "Apache-2.0",
"peer": true,
"bin": {
"tsc": "bin/tsc",
"tsserver": "bin/tsserver"
@@ -15956,7 +15938,6 @@
"version": "7.2.6",
"dev": true,
"license": "MIT",
"peer": true,
"dependencies": {
"esbuild": "^0.25.0",
"fdir": "^6.5.0",
@@ -16048,7 +16029,6 @@
"version": "4.0.14",
"dev": true,
"license": "MIT",
"peer": true,
"dependencies": {
"@vitest/expect": "4.0.14",
"@vitest/mocker": "4.0.14",
@@ -16162,7 +16142,6 @@
"version": "5.103.0",
"dev": true,
"license": "MIT",
"peer": true,
"dependencies": {
"@types/eslint-scope": "^3.7.7",
"@types/estree": "^1.0.8",
@@ -16210,7 +16189,6 @@
"version": "5.1.4",
"dev": true,
"license": "MIT",
"peer": true,
"dependencies": {
"@discoveryjs/json-ext": "^0.5.0",
"@webpack-cli/configtest": "^2.1.1",
@@ -16409,7 +16387,6 @@
"node_modules/winston": {
"version": "3.18.3",
"license": "MIT",
"peer": true,
"dependencies": {
"@colors/colors": "^1.6.0",
"@dabh/diagnostics": "^2.0.8",
@@ -16527,7 +16504,6 @@
"node_modules/ws": {
"version": "8.18.3",
"license": "MIT",
"peer": true,
"engines": {
"node": ">=10.0.0"
},
@@ -16717,7 +16693,6 @@
"node_modules/zod": {
"version": "3.25.76",
"license": "MIT",
"peer": true,
"funding": {
"url": "https://github.com/sponsors/colinhacks"
}
@@ -16838,7 +16813,6 @@
"resolved": "https://registry.npmjs.org/@opentelemetry/api/-/api-1.4.1.tgz",
"integrity": "sha512-O2yRJce1GOc6PAy3QxFM4NzFiWzvScDC1/5ihYBL6BUEVdq0XMWN01sppE+H6bBXbaFYipjwFLEWLg5PaSOThA==",
"license": "Apache-2.0",
"peer": true,
"engines": {
"node": ">=8.0.0"
}
@@ -16848,7 +16822,6 @@
"resolved": "https://registry.npmjs.org/@opentelemetry/api-logs/-/api-logs-0.49.1.tgz",
"integrity": "sha512-kaNl/T7WzyMUQHQlVq7q0oV4Kev6+0xFwqzofryC66jgGMacd0QH5TwfpbUwSTby+SdAdprAe5UKMvBw4tKS5Q==",
"license": "Apache-2.0",
"peer": true,
"dependencies": {
"@opentelemetry/api": "^1.0.0"
},
@@ -19624,7 +19597,7 @@
},
"src/puter-js": {
"name": "@heyputer/puter.js",
"version": "2.2.5",
"version": "2.2.10",
"license": "Apache-2.0",
"dependencies": {
"@heyputer/kv.js": "^0.2.1",
src/backend/src/services/ai/image/AIImageGenerationService.ts
+2 -3
View File
@@ -22,7 +22,6 @@ import { ErrorService } from '../../../modules/core/ErrorService.js';
import { Context } from '../../../util/context.js';
import BaseService from '../../BaseService.js';
import { BaseDatabaseAccessService } from '../../database/BaseDatabaseAccessService.js';
import { DB_WRITE } from '../../database/consts.js';
import { DriverService } from '../../drivers/DriverService.js';
import { TypedValue } from '../../drivers/meta/Runtime.js';
import { EventService } from '../../EventService.js';
@@ -30,8 +29,8 @@ import { MeteringService } from '../../MeteringService/MeteringService.js';
import { GeminiImageGenerationProvider } from './providers/GeminiImageGenerationProvider/GeminiImageGenerationProvider.js';
import { OpenAiImageGenerationProvider } from './providers/OpenAiImageGenerationProvider/OpenAiImageGenerationProvider.js';
import { TogetherImageGenerationProvider } from './providers/TogetherImageGenerationProvider/TogetherImageGenerationProvider.js';
import { XAIImageGenerationProvider } from './providers/XAIImageGenerationProvider/XAIImageGenerationProvider.js';
import { IGenerateParams, IImageModel, IImageProvider } from './providers/types.js';
import { XAIImageGenerationProvider } from './providers/XAIImageGenerationProvider/XAIImageGenerationProvider.js';
export class AIImageGenerationService extends BaseService {
@@ -44,7 +43,7 @@ export class AIImageGenerationService extends BaseService {
}
get db (): BaseDatabaseAccessService {
return this.services.get('database').get(DB_WRITE, 'ai-service');
return this.services.get('database').get();
}
get errorService (): ErrorService {
src/backend/src/services/ai/image/providers/GeminiImageGenerationProvider/GeminiImageGenerationProvider.ts
+119 -5
View File
@@ -30,6 +30,12 @@ type GeminiGenerateParams = IGenerateParams & {
input_image_mime_type?: string;
};
interface GeminiUsageMetadata {
promptTokenCount: number;
candidatesTokenCount: number;
thoughtsTokenCount: number;
}
export class GeminiImageGenerationProvider implements IImageProvider {
#meteringService: MeteringService;
#client: GoogleGenAI;
@@ -85,7 +91,8 @@ export class GeminiImageGenerationProvider implements IImageProvider {
const priceKey = `${quality ? `${quality}:` : ''}${ratio.w}x${ratio.h}`;
const priceInCents = selectedModel.costs[priceKey];
if ( priceInCents === undefined ) {
const availableSizes = Object.keys(selectedModel.costs);
const availableSizes = Object.keys(selectedModel.costs)
.filter(key => key !== 'input' && key !== 'output');
throw APIError.create('field_invalid', undefined, {
key: 'size/quality combination',
expected: `one of: ${ availableSizes.join(', ')}`,
@@ -103,8 +110,11 @@ export class GeminiImageGenerationProvider implements IImageProvider {
});
}
const costInMicroCents = priceInCents * 1_000_000;
const usageAllowed = await this.#meteringService.hasEnoughCredits(actor, costInMicroCents);
const estimatedPromptTokenCount = this.#estimatePromptTokenCount(prompt);
const estimatedInputCostInCents = this.#calculateTokenCostInCents(estimatedPromptTokenCount, selectedModel.costs.input);
const estimatedOutputCostInCents = priceInCents;
const estimatedTotalCostInMicroCents = this.#toMicroCents(estimatedInputCostInCents + estimatedOutputCostInCents);
const usageAllowed = await this.#meteringService.hasEnoughCredits(actor, estimatedTotalCostInMicroCents);
if ( ! usageAllowed ) {
throw APIError.create('insufficient_funds');
@@ -116,8 +126,37 @@ export class GeminiImageGenerationProvider implements IImageProvider {
contents,
});
const usageType = `gemini:${selectedModel.id}:${priceKey}`;
this.#meteringService.incrementUsage(actor, usageType, 1, costInMicroCents);
const usage = this.#extractUsageMetadata(response);
const inputTokenCount = usage.promptTokenCount || estimatedPromptTokenCount;
const outputTokenCount = usage.candidatesTokenCount + usage.thoughtsTokenCount;
const inputCostInCents = this.#calculateTokenCostInCents(inputTokenCount, selectedModel.costs.input);
const outputTextCostInCents = this.#calculateTokenCostInCents(outputTokenCount, selectedModel.costs.output);
const outputCostInCents = priceInCents + outputTextCostInCents;
const usagePrefix = `gemini:${selectedModel.id}:${priceKey}`;
this.#meteringService.batchIncrementUsages(actor, [
{
usageType: `${usagePrefix}:input`,
usageAmount: Math.max(inputTokenCount, 1),
costOverride: this.#toMicroCents(inputCostInCents),
},
{
usageType: `${usagePrefix}:output`,
usageAmount: Math.max(outputTokenCount, 1),
costOverride: this.#toMicroCents(outputCostInCents),
},
]);
this.#setResponseCostMetadata({
model: selectedModel.id,
quality,
ratio,
inputCostInCents,
outputCostInCents,
inputTokenCount,
outputTokenCount,
});
const url = this.#extractImageUrl(response);
@@ -144,6 +183,81 @@ export class GeminiImageGenerationProvider implements IImageProvider {
return `Generate a picture of dimensions ${parseInt(`${ratio.w}`)}x${parseInt(`${ratio.h}`)} with the prompt: ${prompt}`;
}
#setResponseCostMetadata ({
model,
quality,
ratio,
inputCostInCents,
outputCostInCents,
inputTokenCount,
outputTokenCount,
}: {
model: string;
quality?: string;
ratio: { w: number; h: number };
inputCostInCents: number;
outputCostInCents: number;
inputTokenCount: number;
outputTokenCount: number;
}) {
const clientDriverCall = Context.get('client_driver_call') as { response_metadata?: Record<string, unknown> } | undefined;
const responseMetadata = clientDriverCall?.response_metadata;
if ( ! responseMetadata ) return;
const totalCostInCents = inputCostInCents + outputCostInCents;
responseMetadata.cost = {
currency: 'usd-cents',
input: inputCostInCents,
output: outputCostInCents,
total: totalCostInCents,
};
responseMetadata.cost_components = {
provider: 'gemini-image-generation',
model,
quality,
ratio: `${ratio.w}x${ratio.h}`,
input_tokens: inputTokenCount,
output_tokens: outputTokenCount,
input_microcents: this.#toMicroCents(inputCostInCents),
output_microcents: this.#toMicroCents(outputCostInCents),
total_microcents: this.#toMicroCents(totalCostInCents),
};
}
#extractUsageMetadata (response: GenerateContentResponse): GeminiUsageMetadata {
const usage = (response as GenerateContentResponse & { usageMetadata?: Record<string, unknown> }).usageMetadata;
return {
promptTokenCount: this.#toSafeCount(usage?.promptTokenCount),
candidatesTokenCount: this.#toSafeCount(usage?.candidatesTokenCount),
thoughtsTokenCount: this.#toSafeCount(usage?.thoughtsTokenCount),
};
}
#estimatePromptTokenCount (prompt: string): number {
const text = prompt.trim();
if ( text.length === 0 ) return 0;
// Same approximation used by chat billing flow.
return Math.max(1, Math.floor(((text.length / 4) + (text.split(/\s+/).length * (4 / 3))) / 2));
}
#calculateTokenCostInCents (tokenCount: number, centsPerMillion?: number): number {
if ( !Number.isFinite(tokenCount) || tokenCount <= 0 ) return 0;
if ( !Number.isFinite(centsPerMillion) || (centsPerMillion ?? 0) <= 0 ) return 0;
return (tokenCount / 1_000_000) * (centsPerMillion as number);
}
#toMicroCents (cents: number): number {
if ( !Number.isFinite(cents) || cents <= 0 ) return 1;
return Math.ceil(cents * 1_000_000);
}
#toSafeCount (value: unknown): number {
if ( typeof value !== 'number' || !Number.isFinite(value) || value < 0 ) return 0;
return Math.floor(value);
}
#extractImageUrl (response: GenerateContentResponse): string | undefined {
const parts = response?.candidates?.[0]?.content?.parts;
if ( ! Array.isArray(parts) ) {
src/backend/src/services/ai/image/providers/GeminiImageGenerationProvider/models.ts
+54 -47
View File
@@ -26,8 +26,10 @@ export const GEMINI_IMAGE_GENERATION_MODELS: IImageModel[] = [
puterId: 'google:google/gemini-2.5-flash-image',
id: 'gemini-2.5-flash-image',
aliases: [
'gemini-2.5-flash-image-preview', 'gemini-2.5-flash-image',
'google/gemini-2.5-flash-image-preview', 'google/gemini-2.5-flash-image',
'gemini-2.5-flash-image-preview',
'gemini-2.5-flash-image',
'google/gemini-2.5-flash-image-preview',
'google/gemini-2.5-flash-image',
'google:google/gemini-2.5-flash-image-preview',
],
@@ -35,18 +37,20 @@ export const GEMINI_IMAGE_GENERATION_MODELS: IImageModel[] = [
version: '1.0',
costs_currency: 'usd-cents',
index_cost_key: '1x1',
index_input_cost_key: 'prompt_tokens',
allowedQualityLevels: [''],
costs: {
'1x1': 3.9, // $0.039 per image
'2x3': 3.9, // $0.039 per image
'3x2': 3.9, // $0.039 per image
'3x4': 3.9, // $0.039 per image
'4x3': 3.9, // $0.039 per image
'4x5': 3.9, // $0.039 per image
'5x4': 3.9, // $0.039 per image
'9x16': 3.9, // $0.039 per image
'16x9': 3.9, // $0.039 per image
'21x9': 3.9, // $0.039 per image
input: 30, // $0.30 per 1M prompt tokens, however google counts them
'1x1': 3.9, // $0.039 per image, just used for extiamte input allowed usage
'2x3': 3.9, // $0.039 per image, just used for extiamte input allowed usage
'3x2': 3.9, // $0.039 per image, just used for extiamte input allowed usage
'3x4': 3.9, // $0.039 per image, just used for extiamte input allowed usage
'4x3': 3.9, // $0.039 per image, just used for extiamte input allowed usage
'4x5': 3.9, // $0.039 per image, just used for extiamte input allowed usage
'5x4': 3.9, // $0.039 per image, just used for extiamte input allowed usage
'9x16': 3.9, // $0.039 per image, just used for extiamte input allowed usage
'16x9': 3.9, // $0.039 per image, just used for extiamte input allowed usage
'21x9': 3.9, // $0.039 per image, just used for extiamte input allowed usage
},
allowedRatios: [
{ w: 1, h: 1 },
@@ -62,16 +66,17 @@ export const GEMINI_IMAGE_GENERATION_MODELS: IImageModel[] = [
],
},
{
puterId: 'google:google/gemini-3-pro-image',
id: 'gemini-3-pro-image',
puterId: 'google:google/gemini-3-pro-image-preview',
id: 'gemini-3-pro-image-preview',
name: 'Gemini 3 Pro Image',
version: '1.0',
costs_currency: 'usd-cents',
index_cost_key: '1K:1x1',
aliases: [
'gemini-3-pro-image-preview', 'gemini-3-pro-image',
'google/gemini-3-pro-image-preview', 'google/gemini-3-pro-image',
'google:google/gemini-3-pro-image-preview',
'gemini-3-pro-image',
'google/gemini-3-pro-image-preview',
'google/gemini-3-pro-image',
'google:google/gemini-3-pro-image',
],
allowedQualityLevels: ['1K', '2K', '4K'],
allowedRatios: [
@@ -87,36 +92,38 @@ export const GEMINI_IMAGE_GENERATION_MODELS: IImageModel[] = [
{ w: 21, h: 9 },
],
costs: {
'1K:1x1': 13.51, // $0.1351 per image
'1K:2x3': 13.51, // $0.1351 per image
'1K:3x2': 13.51, // $0.1351 per image
'1K:3x4': 13.51, // $0.1351 per image
'1K:4x3': 13.51, // $0.1351 per image
'1K:4x5': 13.51, // $0.1351 per image
'1K:5x4': 13.51, // $0.1351 per image
'1K:9x16': 13.51, // $0.1351 per image
'1K:16x9': 13.51, // $0.1351 per image
'1K:21x9': 13.51, // $0.1351 per image
'2K:1x1': 13.51, // $0.1351 per image
'2K:2x3': 13.51, // $0.1351 per image
'2K:3x2': 13.51, // $0.1351 per image
'2K:3x4': 13.51, // $0.1351 per image
'2K:4x3': 13.51, // $0.1351 per image
'2K:4x5': 13.51, // $0.1351 per image
'2K:5x4': 13.51, // $0.1351 per image
'2K:9x16': 13.51, // $0.1351 per image
'2K:16x9': 13.51, // $0.1351 per image
'2K:21x9': 13.51, // $0.1351 per image
'4K:1x1': 24.1, // $0.24 per image
'4K:2x3': 24.1, // $0.24 per image
'4K:3x2': 24.1, // $0.24 per image
'4K:3x4': 24.1, // $0.24 per image
'4K:4x3': 24.1, // $0.24 per image
'4K:4x5': 24.1, // $0.24 per image
'4K:5x4': 24.1, // $0.24 per image
'4K:9x16': 24.1, // $0.24 per image
'4K:16x9': 24.1, // $0.24 per image
'4K:21x9': 24.1, // $0.24 per image
input: 200, // $2.00 per 1M input tokens, however google counts them
output: 1200, // $12.00 per 1M output tokens, however google counts them
'1K:1x1': 13.4, // $0.134 per image
'1K:2x3': 13.4, // $0.134 per image
'1K:3x2': 13.4, // $0.134 per image
'1K:3x4': 13.4, // $0.134 per image
'1K:4x3': 13.4, // $0.134 per image
'1K:4x5': 13.4, // $0.134 per image
'1K:5x4': 13.4, // $0.134 per image
'1K:9x16': 13.4, // $0.134 per image
'1K:16x9': 13.4, // $0.134 per image
'1K:21x9': 13.4, // $0.134 per image
'2K:1x1': 13.4, // $0.134 per image
'2K:2x3': 13.4, // $0.134 per image
'2K:3x2': 13.4, // $0.134 per image
'2K:3x4': 13.4, // $0.134 per image
'2K:4x3': 13.4, // $0.134 per image
'2K:4x5': 13.4, // $0.134 per image
'2K:5x4': 13.4, // $0.134 per image
'2K:9x16': 13.4, // $0.134 per image
'2K:16x9': 13.4, // $0.134 per image
'2K:21x9': 13.4, // $0.134 per image
'4K:1x1': 24, // $0.24 per image
'4K:2x3': 24, // $0.24 per image
'4K:3x2': 24, // $0.24 per image
'4K:3x4': 24, // $0.24 per image
'4K:4x3': 24, // $0.24 per image
'4K:4x5': 24, // $0.24 per image
'4K:5x4': 24, // $0.24 per image
'4K:9x16': 24, // $0.24 per image
'4K:16x9': 24, // $0.24 per image
'4K:21x9': 24, // $0.24 per image
},
},
];
src/backend/src/services/ai/image/providers/OpenAiImageGenerationProvider/OpenAiImageGenerationProvider.ts
+286 -14
View File
@@ -18,13 +18,24 @@
*/
import openai, { OpenAI } from 'openai';
import { ImageGenerateParamsNonStreaming } from 'openai/resources/images.js';
import { ImageGenerateParamsNonStreaming, ImagesResponse } from 'openai/resources/images.js';
import APIError from '../../../../../api/APIError.js';
import { ErrorService } from '../../../../../modules/core/ErrorService.js';
import { Context } from '../../../../../util/context.js';
import { MeteringService } from '../../../../MeteringService/MeteringService.js';
import { IGenerateParams, IImageProvider } from '../types.js';
import { IGenerateParams, IImageModel, IImageProvider } from '../types.js';
import { OPEN_AI_IMAGE_GENERATION_MODELS } from './models.js';
interface OpenAIImageUsage {
inputTokens: number;
outputTokens: number;
inputTextTokens: number;
inputImageTokens: number;
cachedInputTokens: number;
cachedInputTextTokens: number;
cachedInputImageTokens: number;
}
/**
* Service class for generating images using OpenAI's DALL-E API.
* Extends BaseService to provide image generation capabilities through
@@ -37,6 +48,15 @@ export class OpenAiImageGenerationProvider implements IImageProvider {
#openai: OpenAI;
#errors: ErrorService;
static #NON_SIZE_COST_KEYS = [
'text_input',
'text_cached_input',
'text_output',
'image_input',
'image_cached_input',
'image_output',
];
constructor (config: { apiKey: string }, meteringService: MeteringService, errorService: ErrorService) {
this.#meteringService = meteringService;
this.#openai = new openai.OpenAI({
@@ -81,8 +101,10 @@ export class OpenAiImageGenerationProvider implements IImageProvider {
const size = `${ratio.w}x${ratio.h}`;
const price_key = this.#buildPriceKey(selectedModel.id, quality!, size);
if ( ! selectedModel?.costs[price_key] ) {
const availableSizes = Object.keys(selectedModel?.costs);
const outputPriceInCents = selectedModel?.costs[price_key];
if ( outputPriceInCents === undefined ) {
const availableSizes = Object.keys(selectedModel?.costs)
.filter(key => !OpenAiImageGenerationProvider.#NON_SIZE_COST_KEYS.includes(key));
throw APIError.create('field_invalid', undefined, {
key: 'size/quality combination',
expected: `one of: ${ availableSizes.join(', ')}`,
@@ -100,8 +122,18 @@ export class OpenAiImageGenerationProvider implements IImageProvider {
});
}
const costInMicroCents = selectedModel.costs[price_key] * 1_000_000;
const usageAllowed = await this.#meteringService.hasEnoughCredits(actor, costInMicroCents);
const estimatedPromptTokenCount = this.#estimatePromptTokenCount(prompt);
const estimatedInputCostInCents = this.#calculateInputCostInCents(selectedModel, {
inputTokens: estimatedPromptTokenCount,
inputTextTokens: estimatedPromptTokenCount,
inputImageTokens: 0,
cachedInputTokens: 0,
cachedInputTextTokens: 0,
cachedInputImageTokens: 0,
} as OpenAIImageUsage);
const estimatedOutputCostInCents = outputPriceInCents;
const estimatedTotalCostInMicroCents = this.#toMicroCents(estimatedInputCostInCents + estimatedOutputCostInCents);
const usageAllowed = await this.#meteringService.hasEnoughCredits(actor, estimatedTotalCostInMicroCents);
if ( ! usageAllowed ) {
throw APIError.create('insufficient_funds');
@@ -117,18 +149,55 @@ export class OpenAiImageGenerationProvider implements IImageProvider {
const result = await this.#openai.images.generate(apiParams);
// For image generation, usage is typically image count and resolution
const usageType = `openai:${selectedModel.id}:${price_key}`;
this.#meteringService.incrementUsage(actor, usageType, 1, costInMicroCents);
const usage = this.#extractUsage(result);
const hasInputTokenUsage =
usage.inputTokens > 0 ||
usage.inputTextTokens > 0 ||
usage.inputImageTokens > 0;
const hasOutputTokenUsage = usage.outputTokens > 0;
const spending_meta = {
model,
size: `${ratio.w}x${ratio.h}`,
const billableUsage = hasInputTokenUsage ? usage : {
...usage,
inputTokens: estimatedPromptTokenCount,
inputTextTokens: estimatedPromptTokenCount,
};
if ( quality ) {
spending_meta.size = `${quality}:${ spending_meta.size}`;
const inputCostInCents = hasInputTokenUsage
? this.#calculateInputCostInCents(selectedModel, billableUsage)
: estimatedInputCostInCents;
const outputCostInCents = this.#calculateOutputCostInCents(selectedModel, usage, outputPriceInCents);
const usageType = `openai:${selectedModel.id}:${price_key}`;
const usageEntries: Array<{ usageType: string; usageAmount: number; costOverride: number }> = [];
if ( inputCostInCents > 0 ) {
usageEntries.push({
usageType: `${usageType}:input`,
usageAmount: Math.max(billableUsage.inputTokens || estimatedPromptTokenCount, 1),
costOverride: this.#toMicroCents(inputCostInCents),
});
}
if ( outputCostInCents > 0 ) {
usageEntries.push({
usageType: `${usageType}:output`,
usageAmount: Math.max(usage.outputTokens, 1),
costOverride: this.#toMicroCents(outputCostInCents),
});
}
if ( usageEntries.length ) {
this.#meteringService.batchIncrementUsages(actor, usageEntries);
}
this.#setResponseCostMetadata({
model: selectedModel.id,
quality,
ratio,
inputCostInCents,
outputCostInCents,
usage: billableUsage,
inputUsageSource: hasInputTokenUsage ? 'token-usage' : 'prompt-estimate',
outputUsageSource: hasOutputTokenUsage ? 'token-usage' : 'per-image-fallback',
outputPriceInCents,
});
const url = result.data?.[0]?.url || (result.data?.[0]?.b64_json ? `data:image/png;base64,${ result.data[0].b64_json}` : null);
@@ -139,6 +208,209 @@ export class OpenAiImageGenerationProvider implements IImageProvider {
return url;
}
#extractUsage (result: ImagesResponse): OpenAIImageUsage {
const usage = (result.usage ?? {}) as ImagesResponse.Usage & Record<string, unknown>;
const inputTokens = this.#toSafeCount(usage.input_tokens);
const outputTokens = this.#toSafeCount(usage.output_tokens);
const inputDetails = (usage.input_tokens_details ?? {}) as unknown as Record<string, unknown>;
const inputTextTokens = this.#toSafeCount(inputDetails.text_tokens);
const inputImageTokens = this.#toSafeCount(inputDetails.image_tokens);
const cachedInputTokens = Math.max(
this.#toSafeCount((usage as Record<string, unknown>).cached_input_tokens),
this.#toSafeCount(inputDetails.cached_tokens),
);
const cachedDetails = ((inputDetails.cached_tokens_details || inputDetails.cache_tokens_details) ?? {}) as Record<string, unknown>;
const cachedInputTextTokens = this.#toSafeCount(cachedDetails.text_tokens);
const cachedInputImageTokens = this.#toSafeCount(cachedDetails.image_tokens);
return {
inputTokens,
outputTokens,
inputTextTokens,
inputImageTokens,
cachedInputTokens,
cachedInputTextTokens,
cachedInputImageTokens,
};
}
#calculateInputCostInCents (selectedModel: IImageModel, usage: OpenAIImageUsage): number {
if ( ! this.#isGptImageModel(selectedModel.id) ) {
return 0;
}
const textInputRate = this.#getCostRate(selectedModel, 'text_input');
const textCachedInputRate = this.#getCostRate(selectedModel, 'text_cached_input') ?? textInputRate;
const imageInputRate = this.#getCostRate(selectedModel, 'image_input');
const imageCachedInputRate = this.#getCostRate(selectedModel, 'image_cached_input') ?? imageInputRate;
if ( textInputRate === undefined && imageInputRate === undefined ) {
return 0;
}
const totalInputTokens = Math.max(usage.inputTokens, usage.inputTextTokens + usage.inputImageTokens);
let textTokens = usage.inputTextTokens;
let imageTokens = usage.inputImageTokens;
// Current image generate calls are usually text-only prompts.
if ( textTokens + imageTokens === 0 && totalInputTokens > 0 ) {
textTokens = totalInputTokens;
}
const knownInputTokens = textTokens + imageTokens;
let cachedInputTokens = Math.min(usage.cachedInputTokens, knownInputTokens || totalInputTokens);
let cachedTextTokens = Math.min(usage.cachedInputTextTokens, textTokens);
let cachedImageTokens = Math.min(usage.cachedInputImageTokens, imageTokens);
let cachedRemaining = Math.max(0, cachedInputTokens - (cachedTextTokens + cachedImageTokens));
if ( cachedRemaining > 0 ) {
const availableText = Math.max(textTokens - cachedTextTokens, 0);
const availableImage = Math.max(imageTokens - cachedImageTokens, 0);
const availableTotal = availableText + availableImage;
if ( availableTotal > 0 ) {
const proportionalText = Math.min(availableText, Math.round((availableText / availableTotal) * cachedRemaining));
cachedTextTokens += proportionalText;
cachedRemaining -= proportionalText;
const proportionalImage = Math.min(availableImage, cachedRemaining);
cachedImageTokens += proportionalImage;
cachedRemaining -= proportionalImage;
}
if ( cachedRemaining > 0 && textTokens > cachedTextTokens ) {
const extraText = Math.min(textTokens - cachedTextTokens, cachedRemaining);
cachedTextTokens += extraText;
cachedRemaining -= extraText;
}
if ( cachedRemaining > 0 && imageTokens > cachedImageTokens ) {
const extraImage = Math.min(imageTokens - cachedImageTokens, cachedRemaining);
cachedImageTokens += extraImage;
cachedRemaining -= extraImage;
}
}
const uncachedTextTokens = Math.max(textTokens - cachedTextTokens, 0);
const uncachedImageTokens = Math.max(imageTokens - cachedImageTokens, 0);
return this.#costForTokens(uncachedTextTokens, textInputRate)
+ this.#costForTokens(cachedTextTokens, textCachedInputRate)
+ this.#costForTokens(uncachedImageTokens, imageInputRate)
+ this.#costForTokens(cachedImageTokens, imageCachedInputRate);
}
#calculateOutputCostInCents (selectedModel: IImageModel, usage: OpenAIImageUsage, fallbackPriceInCents: number): number {
if ( ! this.#isGptImageModel(selectedModel.id) ) {
return fallbackPriceInCents;
}
if ( usage.outputTokens <= 0 ) {
return fallbackPriceInCents;
}
const imageOutputRate = this.#getCostRate(selectedModel, 'image_output');
if ( imageOutputRate !== undefined ) {
return this.#costForTokens(usage.outputTokens, imageOutputRate);
}
const textOutputRate = this.#getCostRate(selectedModel, 'text_output');
if ( textOutputRate !== undefined ) {
return this.#costForTokens(usage.outputTokens, textOutputRate);
}
return fallbackPriceInCents;
}
#setResponseCostMetadata ({
model,
quality,
ratio,
inputCostInCents,
outputCostInCents,
usage,
inputUsageSource,
outputUsageSource,
outputPriceInCents,
}: {
model: string;
quality?: string;
ratio: { w: number; h: number };
inputCostInCents: number;
outputCostInCents: number;
usage: OpenAIImageUsage;
inputUsageSource: 'token-usage' | 'prompt-estimate';
outputUsageSource: 'token-usage' | 'per-image-fallback';
outputPriceInCents: number;
}) {
const clientDriverCall = Context.get('client_driver_call') as { response_metadata?: Record<string, unknown> } | undefined;
const responseMetadata = clientDriverCall?.response_metadata;
if ( ! responseMetadata ) return;
const totalCostInCents = inputCostInCents + outputCostInCents;
responseMetadata.cost = {
currency: 'usd-cents',
input: inputCostInCents,
output: outputCostInCents,
total: totalCostInCents,
};
responseMetadata.cost_components = {
provider: 'openai-image-generation',
model,
quality,
ratio: `${ratio.w}x${ratio.h}`,
input_usage_source: inputUsageSource,
output_usage_source: outputUsageSource,
output_image_price_cents: outputPriceInCents,
input_tokens: usage.inputTokens,
output_tokens: usage.outputTokens,
input_text_tokens: usage.inputTextTokens,
input_image_tokens: usage.inputImageTokens,
cached_input_tokens: usage.cachedInputTokens,
cached_input_text_tokens: usage.cachedInputTextTokens,
cached_input_image_tokens: usage.cachedInputImageTokens,
input_microcents: this.#toMicroCents(inputCostInCents),
output_microcents: this.#toMicroCents(outputCostInCents),
total_microcents: this.#toMicroCents(totalCostInCents),
};
}
#estimatePromptTokenCount (prompt: string): number {
const text = prompt.trim();
if ( text.length === 0 ) return 0;
// Same approximation used by chat and Gemini image billing flows.
return Math.max(1, Math.floor(((text.length / 4) + (text.split(/\s+/).length * (4 / 3))) / 2));
}
#getCostRate (selectedModel: IImageModel, key: string): number | undefined {
const value = selectedModel.costs[key];
if ( ! Number.isFinite(value) ) {
return undefined;
}
return value;
}
#costForTokens (tokenCount: number, centsPerMillion?: number): number {
if ( !Number.isFinite(tokenCount) || tokenCount <= 0 ) return 0;
if ( !Number.isFinite(centsPerMillion) || (centsPerMillion ?? 0) <= 0 ) return 0;
return (tokenCount / 1_000_000) * (centsPerMillion as number);
}
#toMicroCents (cents: number): number {
if ( !Number.isFinite(cents) || cents <= 0 ) return 1;
return Math.ceil(cents * 1_000_000);
}
#toSafeCount (value: unknown): number {
if ( typeof value !== 'number' || !Number.isFinite(value) || value < 0 ) return 0;
return Math.floor(value);
}
#isGptImageModel (model: string) {
// Covers gpt-image-1, gpt-image-1-mini, gpt-image-1.5 and future variants.
return model.startsWith('gpt-image-1');
src/backend/src/services/ai/image/providers/OpenAiImageGenerationProvider/models.ts
+28 -3
View File
@@ -10,15 +10,24 @@ export const OPEN_AI_IMAGE_GENERATION_MODELS: IImageModel[] = [
costs_currency: 'usd-cents',
index_cost_key: 'low:1024x1024',
costs: {
// Text tokens (per 1M tokens)
text_input: 500, // $5.00
text_cached_input: 125, // $1.25
text_output: 1000, // $10.00
// Image tokens (per 1M tokens)
image_input: 800, // $8.00
image_cached_input: 200, // $2.00
image_output: 3200, // $32.00
// Image generation (per image)
'low:1024x1024': 0.9,
'low:1024x1536': 1.3,
'low:1536x1024': 1.3,
'medium:1024x1024': 3.4,
'medium:1024x1536': 5.1,
'medium:1024x1536': 5,
'medium:1536x1024': 5,
'high:1024x1024': 13.3,
'high:1024x1536': 20,
'high:1536x1024': 19.9,
'high:1536x1024': 20,
},
allowedQualityLevels: ['low', 'medium', 'high'],
allowedRatios: [{ w: 1024, h: 1024 }, { w: 1024, h: 1536 }, { w: 1536, h: 1024 }],
@@ -32,6 +41,14 @@ export const OPEN_AI_IMAGE_GENERATION_MODELS: IImageModel[] = [
costs_currency: 'usd-cents',
index_cost_key: 'low:1024x1024',
costs: {
// Text tokens (per 1M tokens)
text_input: 200, // $2.00
text_cached_input: 20, // $0.20
// Image tokens (per 1M tokens)
image_input: 250, // $2.50
image_cached_input: 25, // $0.25
image_output: 800, // $8.00
// Image generation (per image)
'low:1024x1024': 0.5,
'low:1024x1536': 0.6,
'low:1536x1024': 0.6,
@@ -54,6 +71,14 @@ export const OPEN_AI_IMAGE_GENERATION_MODELS: IImageModel[] = [
costs_currency: 'usd-cents',
index_cost_key: 'low:1024x1024',
costs: {
// Text tokens (per 1M tokens)
text_input: 500, // $5.00
text_cached_input: 125, // $1.25
// Image tokens (per 1M tokens)
image_input: 1000, // $10.00
image_cached_input: 250, // $2.50
image_output: 4000, // $40.00
// Image generation (per image)
'low:1024x1024': 1.1,
'low:1024x1536': 1.6,
'low:1536x1024': 1.6,
@@ -99,6 +124,6 @@ export const OPEN_AI_IMAGE_GENERATION_MODELS: IImageModel[] = [
'512x512': 1.8,
'1024x1024': 2,
},
allowedRatios: [ { w: 256, h: 256 }, { w: 512, h: 512 }, { w: 1024, h: 1024 }],
allowedRatios: [{ w: 256, h: 256 }, { w: 512, h: 512 }, { w: 1024, h: 1024 }],
},
];
src/backend/src/services/ai/image/providers/types.ts
+1
View File
@@ -8,6 +8,7 @@ export interface IImageModel {
version?: string;
costs_currency: string;
index_cost_key?: string;
index_input_cost_key?: string;
costs: Record<string, number>;
allowedQualityLevels?: string[];
allowedRatios?: { w: number, h: number }[];
src/backend/src/util/securehttp.js
-2
View File
@@ -176,8 +176,6 @@ async function secureAxiosRequest (axios, url, options = {}) {
// Validate URL doesn't contain IP addresses
validateUrlNoIP(url);
console.log(`[securehttp] Making secure request to ${url}`);
// Create secure agents
const { httpAgent, httpsAgent } = createSecureAgents();
src/puter-js/package-lock.json
Generated
+2 -2
View File
@@ -1,12 +1,12 @@
{
"name": "puter",
"version": "2.2.5",
"version": "2.2.10",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "puter",
"version": "2.2.5",
"version": "2.2.10",
"license": "Apache-2.0",
"dependencies": {
"@heyputer/kv.js": "^0.1.92",
src/puter-js/package.json
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "@heyputer/puter.js",
"version": "2.2.5",
"version": "2.2.10",
"description": "Puter.js - A JavaScript library for interacting with Puter services.",
"homepage": "https://developer.puter.com",
"main": "src/index.js",