codecortex-context-engine 0.1.0a7__tar.gz → 0.1.0a8__tar.gz

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/.gitattributes

@@ -14,8 +14,6 @@
 #     the code it describes.
 #
 # `-text` disables all end-of-line conversion in both directions.
-tests/fixtures/upstream/** -text
-tests/fixtures/real_index/** -text
 
-# A serialized index is protobuf, not text, in any location.
-*.scip binary
+# Precision index payloads are binary in any location.
+*.cortexidx binary

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/.github/workflows/bootstrap-release.yml

@@ -18,19 +18,19 @@ jobs:
         with:
           fetch-depth: 0
           persist-credentials: true
-      - name: Create v0.1.0a7 release tag from validated main
+      - name: Create v0.1.0a8 release tag from validated main
         shell: bash
         run: |
           set -euo pipefail
           git config user.name "Behnam Jalali"
           git config user.email "behnamjalalico@gmail.com"
-          if git rev-parse -q --verify refs/tags/v0.1.0a7 >/dev/null; then
-            echo "v0.1.0a7 already exists"
+          if git rev-parse -q --verify refs/tags/v0.1.0a8 >/dev/null; then
+            echo "v0.1.0a8 already exists"
           else
-            git tag -a v0.1.0a7 -m "CodeCortex Context Engine v0.1.0a7"
-            git push origin refs/tags/v0.1.0a7
+            git tag -a v0.1.0a8 -m "CodeCortex Context Engine v0.1.0a8"
+            git push origin refs/tags/v0.1.0a8
           fi
       - name: Dispatch release workflow
         env:
           GH_TOKEN: ${{ github.token }}
-        run: gh workflow run release.yml --ref main -f tag=v0.1.0a7
+        run: gh workflow run release.yml --ref main -f tag=v0.1.0a8

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/.github/workflows/ci.yml

@@ -76,7 +76,6 @@ jobs:
           tests/test_precision_intelligence.py
           tests/test_precision_semantics.py
           tests/test_precision_conformance.py
-          tests/test_real_index_conformance.py
           tests/test_precision_generator.py
           tests/test_dependency_intelligence.py
           tests/test_dependency_contract.py

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/.github/workflows/dependency-docs-live.yml

@@ -19,6 +19,7 @@ jobs:
     timeout-minutes: 10
     env:
       CODECORTEX_DEPENDENCY_DOCS_API_KEY: ${{ secrets.CODECORTEX_DEPENDENCY_DOCS_API_KEY }}
+      CODECORTEX_DEPENDENCY_DOCS_BASE_URL: ${{ vars.CODECORTEX_DEPENDENCY_DOCS_BASE_URL }}
     steps:
       - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
         with:
@@ -33,7 +34,7 @@ jobs:
         id: credential
         shell: bash
         run: |
-          if [ -n "${CODECORTEX_DEPENDENCY_DOCS_API_KEY:-}" ]; then
+          if [ -n "${CODECORTEX_DEPENDENCY_DOCS_API_KEY:-}" ] && [ -n "${CODECORTEX_DEPENDENCY_DOCS_BASE_URL:-}" ]; then
             echo "available=true" >> "$GITHUB_OUTPUT"
           else
             echo "available=false" >> "$GITHUB_OUTPUT"
@@ -44,4 +45,4 @@ jobs:
       - name: Record credential-gated skip
         if: steps.credential.outputs.available != 'true'
         run: |
-          echo "SKIPPED — CODECORTEX_DEPENDENCY_DOCS_API_KEY unavailable" | tee -a "$GITHUB_STEP_SUMMARY"
+          echo "SKIPPED — dependency documentation provider URL or credential unavailable" | tee -a "$GITHUB_STEP_SUMMARY"

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/.github/workflows/structural-engine.yml

@@ -1,9 +1,5 @@
 name: Structural Engine
 
-# The structural engine is an optional extra. Core CI runs the structural tests
-# against an in-repository stub, so this workflow is the one place the real,
-# version-pinned binary is exercised. No credentials are required.
-
 on:
   workflow_dispatch:
   push:
@@ -12,14 +8,12 @@ on:
       - "src/codecortex/structural/**"
       - "tests/test_structural_intelligence.py"
       - "tests/fixtures/structural_engine.py"
-      - "pyproject.toml"
       - ".github/workflows/structural-engine.yml"
   pull_request:
     paths:
       - "src/codecortex/structural/**"
       - "tests/test_structural_intelligence.py"
       - "tests/fixtures/structural_engine.py"
-      - "pyproject.toml"
       - ".github/workflows/structural-engine.yml"
 
 permissions:
@@ -41,21 +35,6 @@ jobs:
           python-version: ${{ matrix.python-version }}
           cache: pip
       - run: python -m pip install --upgrade pip
-      - run: pip install -e ".[dev,structural]"
-      - name: Confirm the pinned engine is installed
-        run: |
-          ast-grep --version
-          ast-grep --version | grep -qF "$(python -c 'from codecortex.structural.engine import TESTED_ENGINE_VERSION as v; print(v)')"
-      - name: Structural search, preview, and guarded apply
+      - run: pip install -e ".[dev]"
+      - name: Structural adapter and guarded rewrite tests
         run: pytest -q tests/test_structural_intelligence.py tests/test_evidence_benchmark.py
-      - name: Real-binary conformance
-        run: pytest -q tests/test_structural_conformance.py
-      - name: Measured evidence benchmark
-        run: |
-          cortex evidence-benchmark --output benchmarks/evidence.json
-      - uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
-        if: always()
-        with:
-          name: evidence-benchmark-${{ matrix.python-version }}
-          path: benchmarks/evidence.json
-          if-no-files-found: warn

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/CHANGELOG.md

@@ -2,6 +2,29 @@
 
 All notable changes to CodeCortex are documented here. The project follows semantic versioning while in alpha; breaking changes may occur before 1.0 and will be called out explicitly.
 
+## 0.1.0a8
+
+- Removed obsolete named upstream attribution/provenance records from the current CodeCortex tree.
+- Removed vendored/vendor-specific conformance fixtures and replaced them with repository-owned invariants.
+- Made optional documentation and structural integrations explicitly configurable and vendor-neutral.
+- Switched precision index discovery to CodeCortex-native `.cortexidx` paths.
+- Preserved the native parser sandbox/discovery hardening already merged to `main`.
+- Revalidated CI, security, Windows compatibility, packaging, Docker, and MCP release paths after the cleanup.
+
+## Unreleased (fixes)
+
+- Native Tree-sitter parsing now runs in a supervised worker process, so a grammar
+  that faults can no longer take down `cortex index` with SIGSEGV. A crashed or
+  hung worker degrades that one file to the fallback parser and indexing
+  continues; `CODECORTEX_NATIVE_INPROCESS=1` restores in-process parsing.
+- The native provider keeps one parser per language instead of building a new
+  parser (and language) for every file.
+- Repository discovery honours `.gitignore` inside a Git work tree, so ignored
+  build output and nested checkouts are no longer indexed as project code.
+- The structural engine is now found when it was installed by the `structural`
+  extra into an isolated environment (`uv tool install`, `pipx`) whose script
+  directory is not on `PATH`.
+
 ## 0.1.0a7
 
 - Fixed a Windows lock-directory race in `FileMutex` discovered by the release matrix.
@@ -37,7 +60,7 @@ All notable changes to CodeCortex are documented here. The project follows seman
 - New MCP tools: `cortex_precise_definition`, `cortex_precise_references`, `cortex_precise_implementations`, `cortex_symbol_occurrences`, `cortex_precision_status`, `cortex_dependency_info`, `cortex_dependency_docs`, `cortex_dependency_context`, `cortex_structural_search`, `cortex_rewrite_preview`, and `cortex_rewrite_apply` (mutating surface only).
 - New CLI commands: `definition`, `references`, `implementations`, `precision-status`, `dependency`, `dependency-docs`, `structural-search`, `rewrite-preview`, `rewrite-apply`, and `evidence-benchmark`.
 - Measured heuristic-versus-evidence benchmark cases for duplicate symbol names, resolved dependency versions, and mechanical migrations; unmeasurable strategies are reported as skipped.
-- `docs/EVIDENCE_FUSION.md`, plus provenance records for the integrated protocol and optional dependencies under `docs/provenance/`.
+- `docs/EVIDENCE_FUSION.md` documents the shipped evidence-fusion surfaces and fallback behavior.
 
 ### Changed
 

codecortex_context_engine-0.1.0a8/HARDENING_REPORT.md

@@ -0,0 +1,41 @@
+# CodeCortex Hardening Report
+
+This report records CodeCortex-owned hardening work and verification evidence.
+
+## Scope
+
+The hardening pass focused on:
+
+- precision navigation and range handling;
+- repository-scoped symbol resolution;
+- position-encoding correctness;
+- path and import security;
+- staleness detection;
+- dependency documentation transport hardening;
+- structural search and guarded rewrite boundaries;
+- evidence ranking and freshness behavior;
+- release, security, and regression coverage.
+
+## Verification
+
+The repository enforces:
+
+- Ruff linting;
+- strict mypy checks;
+- pytest across supported Python versions;
+- coverage threshold at or above 90%;
+- CodeQL and security workflows;
+- package build and install smoke tests;
+- Windows compatibility checks;
+- Docker build smoke tests;
+- signed release artifacts, checksums, SBOM, and provenance.
+
+## Evidence principles
+
+CodeCortex reports measured behavior only when it has a reproducible repository-owned test or benchmark. Optional capabilities that are unavailable are reported as unavailable rather than silently simulated.
+
+## Release discipline
+
+A release is considered valid only when the release matrix, package build, signing, publication, and configured distribution checks complete successfully for the exact release commit.
+
+For current status, use the GitHub Actions results and release artifacts attached to the latest published release.

codecortex_context_engine-0.1.0a8/NOTICE

@@ -0,0 +1,4 @@
+CodeCortex Context Engine
+Copyright 2026 Behnam Jalali
+
+CodeCortex is distributed under the Apache License, Version 2.0. See LICENSE.

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: codecortex-context-engine
-Version: 0.1.0a7
+Version: 0.1.0a8
 Summary: Context intelligence infrastructure for AI coding agents
 Project-URL: Homepage, https://github.com/BehnamJalaliCo/CodeCortex
 Project-URL: Repository, https://github.com/BehnamJalaliCo/CodeCortex
@@ -239,8 +239,6 @@ Requires-Dist: cyclonedx-bom>=4; extra == 'security'
 Requires-Dist: pip-audit>=2.7; extra == 'security'
 Provides-Extra: semantic
 Requires-Dist: sentence-transformers<6,>=3; extra == 'semantic'
-Provides-Extra: structural
-Requires-Dist: ast-grep-cli==0.45.3; extra == 'structural'
 Provides-Extra: web
 Requires-Dist: fastapi<1,>=0.116; extra == 'web'
 Requires-Dist: uvicorn<1,>=0.35; extra == 'web'
@@ -1117,7 +1115,6 @@ Public interfaces are still evolving. Breaking changes can occur before 1.0. Eva
 - [Distributed operation](docs/DISTRIBUTED.md)
 - [Advanced intelligence](docs/ADVANCED_INTELLIGENCE.md)
 - [Evidence fusion](docs/EVIDENCE_FUSION.md)
-- [Provenance records](docs/provenance/)
 - [Integrations](docs/INTEGRATIONS.md)
 - [Quality](docs/QUALITY.md)
 - [Testing](docs/TESTING.md)
@@ -1128,7 +1125,6 @@ Public interfaces are still evolving. Breaking changes can occur before 1.0. Eva
 - [Governance](GOVERNANCE.md)
 - [Roadmap](ROADMAP.md)
 
-Third-party license and provenance obligations remain in the repository's legal/provenance files. Product-facing documentation uses CodeCortex-native capability names.
 
 ---
 
@@ -1950,7 +1946,6 @@ interfaceهای عمومی هنوز تکامل پیدا می‌کنند و قب
 - [Distributed](docs/DISTRIBUTED.md)
 - [Advanced Intelligence](docs/ADVANCED_INTELLIGENCE.md)
 - [Evidence Fusion](docs/EVIDENCE_FUSION.md)
-- [Provenance](docs/provenance/)
 - [Integrations](docs/INTEGRATIONS.md)
 - [Quality](docs/QUALITY.md)
 - [Testing](docs/TESTING.md)
@@ -1961,7 +1956,6 @@ interfaceهای عمومی هنوز تکامل پیدا می‌کنند و قب
 - [Governance](GOVERNANCE.md)
 - [Roadmap](ROADMAP.md)
 
-تعهدات legal و provenance اجزای ثالث در فایل‌های حقوقی ریپو باقی می‌مانند. متن product-facing قابلیت‌ها را با زبان خود CodeCortex توضیح می‌دهد.
 
 ---
 

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/README.md

@@ -869,7 +869,6 @@ Public interfaces are still evolving. Breaking changes can occur before 1.0. Eva
 - [Distributed operation](docs/DISTRIBUTED.md)
 - [Advanced intelligence](docs/ADVANCED_INTELLIGENCE.md)
 - [Evidence fusion](docs/EVIDENCE_FUSION.md)
-- [Provenance records](docs/provenance/)
 - [Integrations](docs/INTEGRATIONS.md)
 - [Quality](docs/QUALITY.md)
 - [Testing](docs/TESTING.md)
@@ -880,7 +879,6 @@ Public interfaces are still evolving. Breaking changes can occur before 1.0. Eva
 - [Governance](GOVERNANCE.md)
 - [Roadmap](ROADMAP.md)
 
-Third-party license and provenance obligations remain in the repository's legal/provenance files. Product-facing documentation uses CodeCortex-native capability names.
 
 ---
 
@@ -1702,7 +1700,6 @@ interfaceهای عمومی هنوز تکامل پیدا می‌کنند و قب
 - [Distributed](docs/DISTRIBUTED.md)
 - [Advanced Intelligence](docs/ADVANCED_INTELLIGENCE.md)
 - [Evidence Fusion](docs/EVIDENCE_FUSION.md)
-- [Provenance](docs/provenance/)
 - [Integrations](docs/INTEGRATIONS.md)
 - [Quality](docs/QUALITY.md)
 - [Testing](docs/TESTING.md)
@@ -1713,7 +1710,6 @@ interfaceهای عمومی هنوز تکامل پیدا می‌کنند و قب
 - [Governance](GOVERNANCE.md)
 - [Roadmap](ROADMAP.md)
 
-تعهدات legal و provenance اجزای ثالث در فایل‌های حقوقی ریپو باقی می‌مانند. متن product-facing قابلیت‌ها را با زبان خود CodeCortex توضیح می‌دهد.
 
 ---
 

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/docs/EVIDENCE_FUSION.md

@@ -218,7 +218,7 @@ to omit.
 {
   "precision_index": {
     "enabled": true,
-    "path": ".codecortex/precision/index.scip",
+    "path": ".codecortex/precision/index.cortexidx",
     "auto_generate": false,
     "generator_command": []
   },
@@ -241,8 +241,8 @@ to omit.
 ```
 
 A precision index is discovered automatically at
-`.codecortex/precision/index.scip`, `index.scip`, `.scip/index.scip`, or
-`build/index.scip` when no path is configured. A malformed section falls back
+`.codecortex/precision/index.cortexidx`, `index.cortexidx`, `.codecortex/index.cortexidx`, or
+`build/index.cortexidx` when no path is configured. A malformed section falls back
 to defaults rather than breaking local operation.
 
 ## Diagnostics

codecortex_context_engine-0.1.0a8/docs/LICENSING.md

@@ -0,0 +1,25 @@
+# Licensing
+
+## CodeCortex Context Engine
+
+CodeCortex Context Engine is distributed under the Apache License, Version 2.0.
+
+The repository `LICENSE` file is the authoritative license text for CodeCortex-covered material.
+
+Copyright:
+
+- Copyright 2026 Behnam Jalali
+
+Apache-2.0 permits use, reproduction, modification, distribution, sublicensing, and commercial use subject to its terms and conditions.
+
+## Dependencies and release artifacts
+
+Runtime and optional dependencies are declared in `pyproject.toml`. Published releases include a CycloneDX SBOM, cryptographic checksums, signed artifacts, and provenance attestations so consumers can inspect the exact dependency graph of a release.
+
+## Commercial arrangements
+
+The open-source license does not prevent separate commercial support, hosted services, enterprise terms, warranties, or separately licensed material where applicable.
+
+## Trademarks
+
+Copyright licensing and trademark rights are separate. The CodeCortex name and branding are not granted for unrestricted trademark use merely by receiving the source code.

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "codecortex-context-engine"
-version = "0.1.0a7"
+version = "0.1.0a8"
 description = "Context intelligence infrastructure for AI coding agents"
 readme = "README.md"
 requires-python = ">=3.11"
@@ -29,13 +29,6 @@ dependencies = [
 semantic = ["sentence-transformers>=3,<6"]
 parsers = ["tree-sitter-language-pack>=1.15.8,<2"]
 web = ["fastapi>=0.116,<1", "uvicorn>=0.35,<1"]
-# Pinned exactly, not as a range. CodeCortex parses this engine's JSON stream
-# records, so its output shape is part of the integration contract rather than
-# an implementation detail. 0.45.3 is the version the conformance tests in
-# tests/test_structural_conformance.py run against; a range would let an
-# untested release change the record shape silently. `cortex doctor` reports a
-# mismatch when a different version is installed anyway.
-structural = ["ast-grep-cli==0.45.3"]
 dev = [
   "pytest>=8.2",
   "pytest-asyncio>=1.4.0",

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/scripts/generate_readme.py

@@ -303,7 +303,7 @@ See [HARDENING_REPORT.md](HARDENING_REPORT.md) and [benchmarks/](benchmarks/) fo
 OUTRO = r"""
 ## Maintainer and project ownership
 
-CodeCortex is maintained by **Behnam Jalali**. CodeCortex-owned material in this repository is licensed under Apache-2.0. Third-party material remains subject to its applicable copyright and license terms; see `LICENSE`, `NOTICE`, `THIRD_PARTY_NOTICES.md`, and `docs/LICENSING.md` for the repository's licensing records.
+CodeCortex is maintained by **Behnam Jalali**. CodeCortex material in this repository is licensed under Apache-2.0; see `LICENSE`, `NOTICE`, and `docs/LICENSING.md`.
 
 ## Contributing
 

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/server.json

@@ -8,13 +8,13 @@
     "source": "github"
   },
   "websiteUrl": "https://behnamjalalico.github.io/CodeCortex/",
-  "version": "0.1.0a7",
+  "version": "0.1.0a8",
   "packages": [
     {
       "registryType": "pypi",
       "registryBaseUrl": "https://pypi.org",
       "identifier": "codecortex-context-engine",
-      "version": "0.1.0a7",
+      "version": "0.1.0a8",
       "transport": {
         "type": "stdio"
       },

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/src/codecortex/config.py

@@ -45,7 +45,7 @@ class DependencyDocsConfig(BaseModel):
 
     enabled: bool = False
     provider: str = "remote"
-    base_url: str = "https://context7.com/api"
+    base_url: str = ""
     api_key_env: str = "CODECORTEX_DEPENDENCY_DOCS_API_KEY"
     connect_timeout_seconds: float = Field(default=5.0, gt=0)
     read_timeout_seconds: float = Field(default=20.0, gt=0)

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/src/codecortex/dependencies/service.py

@@ -28,6 +28,7 @@ class DependencyDocsStatus:
     """Capability report for the documentation provider."""
 
     enabled: bool
+    endpoint_configured: bool
     provider: str
     credentials_present: bool
     cache_writable: bool
@@ -37,6 +38,8 @@ class DependencyDocsStatus:
     def label(self) -> str:
         if not self.enabled:
             return "disabled"
+        if not self.endpoint_configured:
+            return "not configured"
         if not self.credentials_present:
             return "credentials missing"
         return "available"
@@ -45,6 +48,7 @@ class DependencyDocsStatus:
         return {
             "status": self.label,
             "enabled": self.enabled,
+            "endpoint_configured": self.endpoint_configured,
             "provider": self.provider,
             "credentials_present": self.credentials_present,
             "cache_writable": self.cache_writable,
@@ -139,7 +143,7 @@ class DependencyIntelligence:
     def provider(self) -> DependencyDocumentationProvider | None:
         if self._provider is not None:
             return self._provider
-        if not self.settings.enabled:
+        if not self.settings.enabled or not self.settings.base_url.strip():
             return None
         self._provider = RemoteDocumentationProvider(self.settings, self.api_key())
         return self._provider
@@ -151,10 +155,13 @@ class DependencyIntelligence:
         detail = ""
         if not self.settings.enabled:
             detail = "dependency documentation is disabled in configuration"
+        elif not self.settings.base_url.strip():
+            detail = "set dependency_docs.base_url to enable documentation lookups"
         elif not credentials:
             detail = f"set {self.settings.api_key_env} to enable documentation lookups"
         return DependencyDocsStatus(
             enabled=self.settings.enabled,
+            endpoint_configured=bool(self.settings.base_url.strip()),
             provider=self.settings.provider,
             credentials_present=credentials,
             cache_writable=self.cache.writable(),

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/src/codecortex/evaluation/evidence_benchmark.py

@@ -8,7 +8,6 @@ metric that cannot be measured is reported as ``None`` rather than filled in.
 from __future__ import annotations
 
 import json
-import shutil
 from dataclasses import asdict, dataclass, field
 from pathlib import Path
 from time import perf_counter
@@ -184,7 +183,7 @@ class EvidenceBenchmark:
             )
             return
 
-        index_path = root / ".codecortex" / "precision" / "index.scip"
+        index_path = root / ".codecortex" / "precision" / "index.cortexidx"
         index_path.parent.mkdir(parents=True, exist_ok=True)
         index_path.write_bytes(self.precision_index)
         started = perf_counter()
@@ -277,7 +276,7 @@ class EvidenceBenchmark:
             duration_ms=(perf_counter() - started) * 1000,
         )
 
-        command = self.structural_command or shutil.which("ast-grep")
+        command = self.structural_command
         if command is None:
             self._skipped.append("mechanical-migration: structural engine is not installed")
             self._cases.append(

codecortex_context_engine-0.1.0a8/src/codecortex/indexing/discovery.py

@@ -0,0 +1,85 @@
+"""Which files on disk count as "the repository".
+
+The walk used to be a bare ``rglob`` filtered by a hard-coded directory set.
+That set cannot know what a given project considers generated, vendored or
+throwaway, so anything the project already declared uninteresting in
+``.gitignore`` was indexed anyway: build output, caches, and — the case that
+made this visible — eight full agent worktrees checked out under a dot
+directory, which turned a 3.5k-file repository into a 15.5k-file one and made
+every impact result list the same symbol repeatedly, once per copy.
+
+Git already answers this question exactly, so when the root is a Git work tree
+its answer is used: tracked files plus untracked-but-not-ignored ones, which is
+also what a contributor sees. Nested checkouts drop out for free because Git
+does not descend into them. Everything else (no Git, no ``git`` binary, a
+timeout) falls back to the previous walk, so behaviour is unchanged off Git.
+"""
+
+from __future__ import annotations
+
+import subprocess
+from pathlib import Path
+
+#: Directories never indexed, whatever Git says about them.
+EXCLUDED_PARTS = frozenset(
+    {".git", ".codecortex", ".venv", "venv", "node_modules", "dist", "build", "__pycache__"}
+)
+
+#: Git may be slow on a cold cache over a huge tree; do not hang the indexer.
+GIT_TIMEOUT_SECONDS = 20.0
+
+
+def is_excluded(relative: Path) -> bool:
+    """True when a repository-relative path lives under an excluded directory."""
+    return any(part in EXCLUDED_PARTS for part in relative.parts)
+
+
+def _git_files(root: Path) -> list[Path] | None:
+    """Files Git would show for ``root``, or None when Git cannot answer."""
+    try:
+        completed = subprocess.run(  # noqa: S603,S607 - fixed argv, no shell
+            [
+                "git",
+                "-C",
+                str(root),
+                "ls-files",
+                "-z",
+                "--cached",
+                "--others",
+                "--exclude-standard",
+            ],
+            capture_output=True,
+            timeout=GIT_TIMEOUT_SECONDS,
+            check=False,
+        )
+    except (OSError, subprocess.SubprocessError):
+        return None
+    if completed.returncode != 0:
+        return None
+    names = completed.stdout.decode("utf-8", errors="replace").split("\0")
+    return [root / name for name in names if name]
+
+
+def iter_repository_files(root: Path, *, max_bytes: int | None = None) -> list[Path]:
+    """Return the repository's files, sorted, honouring ``.gitignore`` when possible."""
+    candidates = _git_files(root)
+    if candidates is None:
+        candidates = [path for path in root.rglob("*")]
+    files: list[Path] = []
+    for path in candidates:
+        try:
+            if not path.is_file():
+                continue
+            relative = path.relative_to(root)
+        except (OSError, ValueError):
+            continue
+        if is_excluded(relative):
+            continue
+        if max_bytes is not None:
+            try:
+                if path.stat().st_size > max_bytes:
+                    continue
+            except OSError:
+                continue
+        files.append(path)
+    return sorted(files)

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/src/codecortex/indexing/incremental.py

@@ -8,9 +8,10 @@ from dataclasses import asdict, dataclass
 from pathlib import Path
 from typing import Any
 
+from codecortex.indexing.discovery import EXCLUDED_PARTS, iter_repository_files
 from codecortex.state import AtomicJsonFile
 
-_EXCLUDED = {".git", ".codecortex", ".venv", "venv", "node_modules", "dist", "build", "__pycache__"}
+_EXCLUDED = EXCLUDED_PARTS
 
 
 @dataclass(frozen=True, slots=True)
@@ -45,19 +46,7 @@ class IncrementalIndex:
         self.max_file_bytes = max_file_bytes
 
     def _iter_files(self) -> list[Path]:
-        files: list[Path] = []
-        for path in self.root.rglob("*"):
-            if not path.is_file():
-                continue
-            relative = path.relative_to(self.root)
-            if any(part in _EXCLUDED for part in relative.parts):
-                continue
-            try:
-                if path.stat().st_size > self.max_file_bytes:
-                    continue
-            except OSError:
-                continue
-            files.append(path)
+        files = iter_repository_files(self.root, max_bytes=self.max_file_bytes)
         files.sort(key=lambda item: item.relative_to(self.root).as_posix())
         return files
 

{codecortex_context_engine-0.1.0a7 → codecortex_context_engine-0.1.0a8}/src/codecortex/indexing/indexer.py

@@ -4,12 +4,13 @@ from __future__ import annotations
 
 from pathlib import Path
 
+from codecortex.indexing.discovery import EXCLUDED_PARTS, iter_repository_files
 from codecortex.indexing.graph import GraphEdge, GraphNode, ProjectGraph
 from codecortex.indexing.relationships import RelationshipExtractor
 from codecortex.indexing.resolution import CrossFileResolver
 from codecortex.languages import LanguageRegistry
 
-_EXCLUDED = {".git", ".codecortex", ".venv", "venv", "node_modules", "dist", "build", "__pycache__"}
+_EXCLUDED = EXCLUDED_PARTS
 
 
 class ProjectIndexer:
@@ -21,17 +22,7 @@ class ProjectIndexer:
         self.resolver = CrossFileResolver()
 
     def _files(self) -> list[Path]:
-        files: list[Path] = []
-        for path in self.root.rglob("*"):
-            if len(files) >= self.max_files:
-                break
-            if not path.is_file():
-                continue
-            relative = path.relative_to(self.root)
-            if any(part in _EXCLUDED for part in relative.parts):
-                continue
-            files.append(path)
-        return sorted(files)
+        return iter_repository_files(self.root)[: self.max_files]
 
     @staticmethod
     def _symbol_id(relative: str, name: str, kind: str, line: int, container: str | None) -> str: