v6: SQL AST extractor + xlsx structural extraction utility (fixes #349)

- extract_sql(): deterministic tree-sitter extraction of tables, views,
  functions, foreign key references, and FROM/JOIN reads_from edges
- .sql added to CODE_EXTENSIONS and dispatch table
- tree-sitter-sql added as optional dep under [sql] extra
- xlsx_extract_structure(): extracts sheet/table/column nodes from .xlsx
  (utility — pipeline wiring in follow-up)
- 6 new SQL tests, 447 total passing

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

graphify/detect.py

@@ -18,7 +18,7 @@ class FileType(str, Enum):
 
 _MANIFEST_PATH = "graphify-out/manifest.json"
 
-CODE_EXTENSIONS = {'.py', '.ts', '.js', '.jsx', '.tsx', '.mjs', '.ejs', '.go', '.rs', '.java', '.cpp', '.cc', '.cxx', '.c', '.h', '.hpp', '.rb', '.swift', '.kt', '.kts', '.cs', '.scala', '.php', '.lua', '.toc', '.zig', '.ps1', '.ex', '.exs', '.m', '.mm', '.jl', '.vue', '.svelte', '.dart', '.v', '.sv'}
+CODE_EXTENSIONS = {'.py', '.ts', '.js', '.jsx', '.tsx', '.mjs', '.ejs', '.go', '.rs', '.java', '.cpp', '.cc', '.cxx', '.c', '.h', '.hpp', '.rb', '.swift', '.kt', '.kts', '.cs', '.scala', '.php', '.lua', '.toc', '.zig', '.ps1', '.ex', '.exs', '.m', '.mm', '.jl', '.vue', '.svelte', '.dart', '.v', '.sv', '.sql'}
 DOC_EXTENSIONS = {'.md', '.mdx', '.txt', '.rst', '.html', '.yaml', '.yml'}
 PAPER_EXTENSIONS = {'.pdf'}
 IMAGE_EXTENSIONS = {'.png', '.jpg', '.jpeg', '.gif', '.webp', '.svg'}
@@ -169,7 +169,6 @@ def xlsx_to_markdown(path: Path) -> str:
             ws = wb[sheet_name]
             rows = []
             for row in ws.iter_rows(values_only=True):
-                # Skip entirely empty rows
                 if all(cell is None for cell in row):
                     continue
                 rows.append([str(cell) if cell is not None else "" for cell in row])
@@ -190,6 +189,91 @@ def xlsx_to_markdown(path: Path) -> str:
         return ""
 
 
+def xlsx_extract_structure(path: Path) -> dict:
+    """Extract structural nodes (sheets, named tables, column headers) from an .xlsx file.
+
+    Returns a nodes/edges dict compatible with the graphify extract pipeline.
+    Used in addition to xlsx_to_markdown so Claude sees both structure and content.
+    """
+    def _nid(*parts: str) -> str:
+        return re.sub(r"[^a-z0-9_]", "_", "_".join(p.lower() for p in parts).strip("_"))
+
+    try:
+        import openpyxl
+    except ImportError:
+        return {"nodes": [], "edges": []}
+
+    try:
+        wb = openpyxl.load_workbook(str(path), read_only=False, data_only=True)
+    except Exception:
+        return {"nodes": [], "edges": []}
+
+    stem = _re.sub(r"[^a-z0-9]", "_", path.stem.lower())
+    str_path = str(path)
+    file_nid = _nid(str_path)
+    nodes: list[dict] = [{"id": file_nid, "label": path.name, "file_type": "document",
+                           "source_file": str_path, "source_location": None}]
+    edges: list[dict] = []
+    seen: set[str] = {file_nid}
+
+    def _add(nid: str, label: str) -> None:
+        if nid not in seen:
+            seen.add(nid)
+            nodes.append({"id": nid, "label": label, "file_type": "document",
+                           "source_file": str_path, "source_location": None})
+
+    def _edge(src: str, tgt: str, relation: str) -> None:
+        edges.append({"source": src, "target": tgt, "relation": relation,
+                       "confidence": "EXTRACTED", "source_file": str_path,
+                       "source_location": None, "weight": 1.0})
+
+    for sheet_name in wb.sheetnames:
+        ws = wb[sheet_name]
+        sheet_nid = _nid(stem, sheet_name)
+        _add(sheet_nid, f"{sheet_name} (sheet)")
+        _edge(file_nid, sheet_nid, "contains")
+
+        # Named Excel Tables (ListObjects)
+        if hasattr(ws, "tables"):
+            for tbl in ws.tables.values():
+                tbl_nid = _nid(stem, sheet_name, tbl.name)
+                _add(tbl_nid, tbl.name)
+                _edge(sheet_nid, tbl_nid, "contains")
+                # Column headers from table header row
+                ref = tbl.ref  # e.g. "A1:D10"
+                if ref:
+                    try:
+                        from openpyxl.utils import range_boundaries
+                        min_col, min_row, max_col, _ = range_boundaries(ref)
+                        header_row = list(ws.iter_rows(min_row=min_row, max_row=min_row,
+                                                       min_col=min_col, max_col=max_col,
+                                                       values_only=True))
+                        if header_row:
+                            for col_name in header_row[0]:
+                                if col_name:
+                                    col_nid = _nid(stem, tbl.name, str(col_name))
+                                    _add(col_nid, str(col_name))
+                                    _edge(tbl_nid, col_nid, "contains")
+                    except Exception:
+                        pass
+        else:
+            # Fallback: first non-empty row as column headers
+            for row in ws.iter_rows(max_row=1, values_only=True):
+                for cell in row:
+                    if cell:
+                        col_nid = _nid(stem, sheet_name, str(cell))
+                        _add(col_nid, str(cell))
+                        _edge(sheet_nid, col_nid, "contains")
+                break
+
+    try:
+        wb.close()
+    except Exception:
+        pass
+
+    return {"nodes": nodes, "edges": edges}
+
+
 def convert_office_file(path: Path, out_dir: Path) -> Path | None:
     """Convert a .docx or .xlsx to a markdown sidecar in out_dir.
 

graphify/extract.py

@@ -1710,6 +1710,137 @@ def extract_verilog(path: Path) -> dict:
     return {"nodes": nodes, "edges": edges}
 
 
+def extract_sql(path: Path) -> dict:
+    """Extract tables, views, functions, and relationships from .sql files via tree-sitter."""
+    try:
+        import tree_sitter_sql as tssql
+        from tree_sitter import Language, Parser
+    except ImportError:
+        return {"nodes": [], "edges": [], "error": "tree_sitter_sql not installed. Run: pip install tree-sitter-sql"}
+
+    try:
+        language = Language(tssql.language())
+        parser = Parser(language)
+        source = path.read_bytes()
+        tree = parser.parse(source)
+        root = tree.root_node
+    except Exception as e:
+        return {"nodes": [], "edges": [], "error": str(e)}
+
+    stem = re.sub(r"[^a-z0-9]", "_", path.stem.lower())
+    str_path = str(path)
+    file_nid = _make_id(str_path)
+    nodes: list[dict] = [{"id": file_nid, "label": path.name, "file_type": "code",
+                           "source_file": str_path, "source_location": None}]
+    edges: list[dict] = []
+    seen_ids: set[str] = {file_nid}
+    table_nids: dict[str, str] = {}  # name → nid for reference resolution
+
+    def _read(n) -> str:
+        return source[n.start_byte:n.end_byte].decode("utf-8", errors="replace")
+
+    def _obj_name(n) -> str | None:
+        for c in n.children:
+            if c.type == "object_reference":
+                for cc in c.children:
+                    if cc.type == "identifier":
+                        return _read(cc)
+        return None
+
+    def _add_node(nid: str, label: str, line: int) -> None:
+        if nid not in seen_ids:
+            seen_ids.add(nid)
+            nodes.append({"id": nid, "label": label, "file_type": "code",
+                           "source_file": str_path, "source_location": f"L{line}"})
+            edges.append({"source": file_nid, "target": nid, "relation": "contains",
+                           "confidence": "EXTRACTED", "source_file": str_path,
+                           "source_location": f"L{line}", "weight": 1.0})
+
+    def _add_edge(src: str, tgt: str, relation: str, line: int) -> None:
+        edges.append({"source": src, "target": tgt, "relation": relation,
+                       "confidence": "EXTRACTED", "source_file": str_path,
+                       "source_location": f"L{line}", "weight": 1.0})
+
+    def walk(node) -> None:
+        t = node.type
+        line = node.start_point[0] + 1
+
+        if t == "create_table":
+            name = _obj_name(node)
+            if name:
+                nid = _make_id(stem, name)
+                _add_node(nid, name, line)
+                table_nids[name.lower()] = nid
+                # Foreign key REFERENCES
+                for col in node.children:
+                    if col.type == "column_definitions":
+                        for cd in col.children:
+                            if cd.type != "column_definition":
+                                continue
+                            ref_name: str | None = None
+                            found_ref = False
+                            for cc in cd.children:
+                                if cc.type == "keyword_references":
+                                    found_ref = True
+                                elif found_ref and cc.type == "object_reference":
+                                    for ccc in cc.children:
+                                        if ccc.type == "identifier":
+                                            ref_name = _read(ccc)
+                                    break
+                            if ref_name:
+                                ref_nid = _make_id(stem, ref_name)
+                                _add_edge(nid, ref_nid, "references", line)
+
+        elif t == "create_view":
+            name = _obj_name(node)
+            if name:
+                nid = _make_id(stem, name)
+                _add_node(nid, name, line)
+                table_nids[name.lower()] = nid
+                # FROM/JOIN table references inside view body
+                _walk_from_refs(node, nid, line)
+
+        elif t == "create_function":
+            name = _obj_name(node)
+            if name:
+                nid = _make_id(stem, name)
+                _add_node(nid, f"{name}()", line)
+                _walk_from_refs(node, nid, line)
+
+        elif t == "create_procedure":
+            name = _obj_name(node)
+            if name:
+                nid = _make_id(stem, name)
+                _add_node(nid, f"{name}()", line)
+                _walk_from_refs(node, nid, line)
+
+        for child in node.children:
+            walk(child)
+
+    def _walk_from_refs(node, caller_nid: str, line: int) -> None:
+        """Recursively find FROM/JOIN table references inside a node."""
+        if node.type in ("from", "join"):
+            for c in node.children:
+                if c.type == "relation":
+                    for cc in c.children:
+                        if cc.type == "object_reference":
+                            for ccc in cc.children:
+                                if ccc.type == "identifier":
+                                    tbl = _read(ccc)
+                                    tbl_nid = _make_id(stem, tbl)
+                                    _add_edge(caller_nid, tbl_nid, "reads_from",
+                                              c.start_point[0] + 1)
+        for child in node.children:
+            _walk_from_refs(child, caller_nid, line)
+
+    for stmt in root.children:
+        if stmt.type == "statement":
+            for child in stmt.children:
+                walk(child)
+
+    return {"nodes": nodes, "edges": edges}
+
+
 def extract_lua(path: Path) -> dict:
     """Extract functions, methods, require() imports, and calls from a .lua file."""
     return _extract_generic(path, _LUA_CONFIG)
@@ -3359,6 +3490,7 @@ def extract(paths: list[Path], cache_root: Path | None = None) -> dict:
         ".dart": extract_dart,
         ".v": extract_verilog,
         ".sv": extract_verilog,
+        ".sql": extract_sql,
     }
 
     total = len(paths)

pyproject.toml

@@ -51,7 +51,8 @@ leiden = ["graspologic; python_version < '3.13'"]
 office = ["python-docx", "openpyxl"]
 video = ["faster-whisper", "yt-dlp"]
 kimi = ["openai"]
-all = ["mcp", "neo4j", "pypdf", "html2text", "watchdog", "graspologic; python_version < '3.13'", "python-docx", "openpyxl", "faster-whisper", "yt-dlp", "matplotlib", "openai"]
+sql = ["tree-sitter-sql"]
+all = ["mcp", "neo4j", "pypdf", "html2text", "watchdog", "graspologic; python_version < '3.13'", "python-docx", "openpyxl", "faster-whisper", "yt-dlp", "matplotlib", "openai", "tree-sitter-sql"]
 
 [project.scripts]
 graphify = "graphify.__main__:main"

tests/fixtures/sample.sql

@@ -0,0 +1,19 @@
+CREATE TABLE organizations (
+  id SERIAL PRIMARY KEY,
+  name TEXT NOT NULL
+);
+
+CREATE TABLE users (
+  id SERIAL PRIMARY KEY,
+  email TEXT NOT NULL,
+  org_id INT REFERENCES organizations(id)
+);
+
+CREATE VIEW active_users AS
+  SELECT * FROM users WHERE active = true;
+
+CREATE FUNCTION get_user(user_id INT) RETURNS users AS $$
+  BEGIN
+    RETURN QUERY SELECT * FROM users WHERE id = user_id;
+  END;
+$$ LANGUAGE plpgsql;

tests/test_multilang.py

@@ -1,9 +1,9 @@
-"""Tests for multi-language AST extraction: JS/TS, Go, Rust."""
+"""Tests for multi-language AST extraction: JS/TS, Go, Rust, SQL."""
 from __future__ import annotations
 import shutil
 from pathlib import Path
 import pytest
-from graphify.extract import extract_js, extract_go, extract_rust, extract
+from graphify.extract import extract_js, extract_go, extract_rust, extract, extract_sql
 
 FIXTURES = Path(__file__).parent / "fixtures"
 
@@ -171,3 +171,38 @@ def test_cache_miss_after_file_change(tmp_path):
     # bar() should appear in the second result
     labels2 = [n["label"] for n in r2["nodes"]]
     assert any("bar" in l for l in labels2)
+
+
+# ── SQL ───────────────────────────────────────────────────────────────────────
+
+def test_sql_finds_tables():
+    r = extract_sql(FIXTURES / "sample.sql")
+    labels = [n["label"] for n in r["nodes"]]
+    assert any("users" in l for l in labels)
+    assert any("organizations" in l for l in labels)
+
+def test_sql_finds_view():
+    r = extract_sql(FIXTURES / "sample.sql")
+    labels = [n["label"] for n in r["nodes"]]
+    assert any("active_users" in l for l in labels)
+
+def test_sql_finds_function():
+    r = extract_sql(FIXTURES / "sample.sql")
+    labels = [n["label"] for n in r["nodes"]]
+    assert any("get_user" in l for l in labels)
+
+def test_sql_emits_foreign_key_edge():
+    r = extract_sql(FIXTURES / "sample.sql")
+    relations = {e["relation"] for e in r["edges"]}
+    assert "references" in relations
+
+def test_sql_emits_reads_from_edge():
+    r = extract_sql(FIXTURES / "sample.sql")
+    relations = {e["relation"] for e in r["edges"]}
+    assert "reads_from" in relations
+
+def test_sql_no_dangling_edges():
+    r = extract_sql(FIXTURES / "sample.sql")
+    node_ids = {n["id"] for n in r["nodes"]}
+    for e in r["edges"]:
+        assert e["source"] in node_ids, f"dangling source: {e['source']}"