@sdsrs/code-graph 0.104.1 → 0.106.0

package/README.md

@@ -367,6 +367,39 @@ lazily on first use. To control that in a restricted environment:
 | `CODE_GRAPH_MODEL_DIR=<dir>` | Load `model.safetensors` from `<dir>` instead of downloading (highest-priority lookup). If the file isn't there, it warns and falls back to the normal search paths. |
 | `CODE_GRAPH_DISABLE_MODEL_DOWNLOAD=1` | Disable the automatic background model download. An already-cached model is still loaded and used; the server otherwise stays in FTS5-only mode. |
 
+### Installing the model by hand
+
+`CODE_GRAPH_MODEL_DIR` is the supported manual route. Hand-populating the
+*default* cache dir (`<platform cache>/code-graph/models`) does **not** work:
+that dir is only trusted once `extract_and_promote` has written a `.model-id`
+marker, so a manually filled copy is treated as not-current and re-downloaded.
+
+```bash
+# Download + verify against the published checksum, then extract flat
+# (the archive has no wrapping folder: model.safetensors, tokenizer.json, config.json).
+curl -sL -o models.tar.gz \
+  https://github.com/sdsrss/code-graph-mcp/releases/download/v<version>/models.tar.gz
+curl -sL https://github.com/sdsrss/code-graph-mcp/releases/download/v<version>/models.tar.gz.sha256
+sha256sum -c models.tar.gz.sha256
+mkdir -p ~/.cache/code-graph/models-manual
+tar -xzf models.tar.gz -C ~/.cache/code-graph/models-manual
+export CODE_GRAPH_MODEL_DIR=~/.cache/code-graph/models-manual
+```
+
+On Windows use `%LOCALAPPDATA%\code-graph\models-manual`, and give `tar` a POSIX
+path for `-C` (`/c/Users/<you>/...`) — `tar -C C:\...` fails with
+"Cannot connect to C: resolve failed". The model is pinned by content hash, so
+a manual install must be redone on each version bump.
+
+### Diagnosing a model that never downloads
+
+`code-graph-mcp doctor` reports the last download outcome, distinguishing
+"no download has been attempted" from "download FAILED after N attempt(s): …"
+with the underlying error. The raw record is
+`<platform cache>/code-graph/model-download.json`. On a network that performs
+TLS inspection, the download retries automatically against the OS certificate
+store after the bundled roots are rejected.
+
 ## Storage
 
 Uses SQLite with:

package/claude-plugin/.claude-plugin/plugin.json

@@ -4,7 +4,7 @@
   "author": {
     "name": "sdsrs"
   },
-  "version": "0.104.1",
+  "version": "0.106.0",
   "keywords": [
     "code-graph",
     "ast",

package/claude-plugin/scripts/doctor.js

@@ -34,8 +34,15 @@ function classifyEmbeddings(hc) {
     return { name: 'Embeddings', status: 'ok', detail: 'no embeddable nodes' };
   }
   if (!done) {
+    // `model_download` carries the LAST recorded download outcome (issue #35).
+    // Without it this printed the same optimistic "retry shortly" on every run,
+    // so a machine whose download can never succeed read as "be patient"
+    // forever instead of "this is broken". Absent field = never attempted,
+    // which is itself a distinct diagnosis, not a reason to advise waiting.
     const why = (hc && hc.embedding_status === 'pending')
-      ? 'model not downloaded/loaded yet — auto-downloads in background on first search; retry shortly or restart the MCP server'
+      ? (hc.model_download
+          ? `last model download: ${hc.model_download}`
+          : 'model not loaded and NO download has ever been attempted on this machine — restart the MCP server, or set CODE_GRAPH_MODEL_DIR to a manually populated model dir (see README → Offline usage)')
       : `embedding_status=${(hc && hc.embedding_status) || 'unknown'}`;
     return { name: 'Embeddings', status: 'warn',
       detail: `vector INACTIVE — ${total} embeddable nodes, 0 embedded; semantic search is FTS5-only (${why})` };

package/claude-plugin/scripts/pr-impact-comment.js

@@ -21,15 +21,46 @@ const MARKER = '<!-- code-graph-impact-review -->';
 const SPAWN_TIMEOUT_MS = 60_000;
 const TOP_AFFECTED = 15;
 
-/// File-level test classifier — mirrors `domain::is_test_path` (Rust). Kept in
-/// sync deliberately; see feedback_test_classifier_dual_sources.md.
+/// Constant lists mirroring `domain::PASCAL_TEST_*` / `INFIX_TEST_EXTS`. Both
+/// sides are generated from their lists rather than transcribed, so adding an
+/// ecosystem is one edit here and one in domain.rs — not a hunt through
+/// hand-written boolean chains.
+const PASCAL_TEST_EXTS = ['cs', 'vb', 'fs', 'java', 'kt', 'scala', 'swift', 'php'];
+const PASCAL_TEST_STEMS = ['Test', 'Tests', 'Spec'];
+const INFIX_TEST_EXTS = ['go', 'rs', 'py', 'dart'];
+
+/// File-level test classifier — mirrors `domain::is_test_path` (Rust). This is
+/// one of several deliberately-synchronized copies; `domain.rs` carries the
+/// "Five sites must agree" note, and two of those sites are intentionally
+/// divergent. Widening only the Rust side breaks the parity contract asserted
+/// by `isTestPath mirrors domain::is_test_path patterns` in the test file, and
+/// makes the "test gaps" section report every Java/C# test as uncovered
+/// production code.
 function isTestPath(p) {
+  // Case-insensitive `test`/`tests` DIRECTORY segment at any depth: xUnit/NUnit
+  // put suites under `src/Tests/<Project>/…` and Maven/Gradle under
+  // `src/test/java/…` (issue #36). Note `toLowerCase()` is Unicode-aware where
+  // Rust uses `to_ascii_lowercase()`; they agree on ASCII paths.
+  const lower = p.toLowerCase();
+  if (
+    lower.startsWith('tests/') || lower.startsWith('test/') ||
+    lower.includes('/tests/') || lower.includes('/test/')
+  ) return true;
+  // PascalCase test-class convention. Case-SENSITIVE and pinned to a known
+  // extension so `src/latest.cs` and `src/mytests.rs` stay production.
+  if (PASCAL_TEST_STEMS.some((stem) => PASCAL_TEST_EXTS.some((ext) => p.endsWith(`${stem}.${ext}`)))) {
+    return true;
+  }
+  if (INFIX_TEST_EXTS.some((ext) => p.endsWith(`_test.${ext}`))) return true;
+  // pytest naming conventions.
+  if (lower.endsWith('.py') &&
+      (lower.startsWith('test_') || lower.includes('/test_') || lower.endsWith('conftest.py'))) {
+    return true;
+  }
   return (
-    p.startsWith('tests/') || p.startsWith('test/') ||
     p.startsWith('benches/') || p.startsWith('bench/') ||
     p.includes('__tests__/') ||
     p.endsWith('/tests.rs') ||
-    p.endsWith('_test.go') || p.endsWith('_test.rs') ||
     p.endsWith('.test.ts') || p.endsWith('.test.js') ||
     p.endsWith('.test.tsx') || p.endsWith('.test.jsx') ||
     p.endsWith('.spec.ts') || p.endsWith('.spec.js') ||

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@sdsrs/code-graph",
-  "version": "0.104.1",
+  "version": "0.106.0",
   "description": "MCP server that indexes codebases into an AST knowledge graph with semantic search, call graph traversal, and HTTP route tracing",
   "license": "MIT",
   "repository": {
@@ -36,10 +36,10 @@
     "node": ">=16"
   },
   "optionalDependencies": {
-    "@sdsrs/code-graph-linux-x64": "0.104.1",
-    "@sdsrs/code-graph-linux-arm64": "0.104.1",
-    "@sdsrs/code-graph-darwin-x64": "0.104.1",
-    "@sdsrs/code-graph-darwin-arm64": "0.104.1",
-    "@sdsrs/code-graph-win32-x64": "0.104.1"
+    "@sdsrs/code-graph-linux-x64": "0.106.0",
+    "@sdsrs/code-graph-linux-arm64": "0.106.0",
+    "@sdsrs/code-graph-darwin-x64": "0.106.0",
+    "@sdsrs/code-graph-darwin-arm64": "0.106.0",
+    "@sdsrs/code-graph-win32-x64": "0.106.0"
   }
 }