coderag-ai 0.3.0__tar.gz → 0.3.2__tar.gz

{coderag_ai-0.3.0 → coderag_ai-0.3.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: coderag-ai
-Version: 0.3.0
+Version: 0.3.2
 Summary: Token-efficient, structure-aware Code Intelligence + RAG for private source repositories.
 Project-URL: Homepage, https://github.com/galacticsurfer/coderag
 Project-URL: Repository, https://github.com/galacticsurfer/coderag
@@ -32,8 +32,8 @@ Requires-Dist: pydantic>=2.7
 Requires-Dist: rich>=13.7
 Requires-Dist: sqlalchemy>=2.0.30
 Requires-Dist: structlog>=24.1
-Requires-Dist: tree-sitter-python<0.22,>=0.21
-Requires-Dist: tree-sitter<0.22,>=0.21
+Requires-Dist: tree-sitter-python>=0.23
+Requires-Dist: tree-sitter>=0.23
 Requires-Dist: typer>=0.12
 Requires-Dist: uvicorn[standard]>=0.30
 Provides-Extra: analyzers
@@ -50,7 +50,7 @@ Provides-Extra: embeddings
 Requires-Dist: sentence-transformers>=3.0; extra == 'embeddings'
 Provides-Extra: localdb
 Requires-Dist: alembic>=1.13; extra == 'localdb'
-Requires-Dist: pgserver>=0.1.4; extra == 'localdb'
+Requires-Dist: pgserver>=0.1.4; (python_version < '3.13') and extra == 'localdb'
 Provides-Extra: mcp
 Requires-Dist: mcp>=2.0; extra == 'mcp'
 Provides-Extra: metrics

{coderag_ai-0.3.0 → coderag_ai-0.3.2}/docs/install-macos.md

@@ -23,6 +23,14 @@ Open a new terminal afterwards so your `PATH` updates.
 
 ### 2. Install CodeRAG
 
+> **Python version:** the `localdb` extra needs **Python 3.9–3.12** (its bundled
+> PostgreSQL has no wheels for 3.13+). On a newer default Python use
+> `pipx install --python python3.12 "coderag-ai[mcp,localdb]"` (`brew install python@3.12`),
+> or install without `localdb` and supply your own PostgreSQL.
+
+> **Corporate network?** If install fails with `invalid peer certificate: UnknownIssuer`,
+> your proxy re-signs TLS. Fix with `export UV_NATIVE_TLS=true` (uses the macOS keychain).
+
 ```bash
 pipx install "coderag-ai[mcp,localdb]"
 ```

{coderag_ai-0.3.0 → coderag_ai-0.3.2}/pyproject.toml

@@ -7,7 +7,7 @@ build-backend = "hatchling.build"
 # command are both `coderag`. Plain "coderag" is taken on PyPI by an unrelated project,
 # and "code-rag" is rejected as too similar (PyPI collapses -/_/. when comparing names).
 name = "coderag-ai"
-version = "0.3.0"
+version = "0.3.2"
 description = "Token-efficient, structure-aware Code Intelligence + RAG for private source repositories."
 readme = "README.md"
 requires-python = ">=3.12"
@@ -40,8 +40,8 @@ dependencies = [
     "alembic>=1.13",
     "psycopg[binary]>=3.1",        # LGPL-3.0 — see docs/licenses.md
     "pgvector>=0.3.0",
-    "tree-sitter>=0.21,<0.22",
-    "tree-sitter-python>=0.21,<0.22",
+    "tree-sitter>=0.23",
+    "tree-sitter-python>=0.23",
     "typer>=0.12",
     "rich>=13.7",
     "httpx>=0.27",
@@ -60,7 +60,10 @@ analyzers = ["pylint>=3.2", "flake8>=7.1"]
 # MCP server: lets Claude Code (and other MCP clients) call CodeRAG retrieval.
 mcp = ["mcp>=2.0"]
 # Rootless local PostgreSQL+pgvector — no Docker, Homebrew, or sudo required.
-localdb = ["pgserver>=0.1.4", "alembic>=1.13"]
+# pgserver only ships wheels for CPython 3.9–3.12, so the marker keeps the extra
+# installable on newer interpreters (the `coderag localdb` command then explains
+# what to do instead of the install failing outright).
+localdb = ["pgserver>=0.1.4; python_version < '3.13'", "alembic>=1.13"]
 # Development / testing.
 dev = [
     "pytest>=8.2",

{coderag_ai-0.3.0 → coderag_ai-0.3.2}/src/coderag/embeddings/sentence_transformer.py

@@ -21,13 +21,18 @@ class SentenceTransformerEmbeddingProvider(EmbeddingProvider):
         except ImportError as exc:  # pragma: no cover - optional dependency
             raise RuntimeError(
                 "sentence-transformers is not installed. Install the 'embeddings' "
-                "extra: pip install 'coderag[embeddings]'"
+                "extra: pip install 'coderag-ai[embeddings]'"
             ) from exc
         self._model = SentenceTransformer(model_name, device=device)
         self.model_name = model_name
         self.model_version = "st1"
         self.batch_size = batch_size
-        self.dimension = int(self._model.get_sentence_embedding_dimension())
+        dim = self._model.get_sentence_embedding_dimension()
+        if dim is None:  # pragma: no cover - model without a fixed dimension
+            raise RuntimeError(
+                f"model {model_name!r} does not report an embedding dimension"
+            )
+        self.dimension = int(dim)
 
     def embed_documents(self, texts: list[str]) -> list[list[float]]:
         vectors = self._model.encode(

{coderag_ai-0.3.0 → coderag_ai-0.3.2}/src/coderag/localdb.py

@@ -19,10 +19,21 @@ def _require_pgserver():
     try:
         import pgserver  # noqa: PLC0415
     except ImportError as exc:  # pragma: no cover - optional dependency
+        import sys
+
+        if sys.version_info >= (3, 13):
+            raise RuntimeError(
+                f"`coderag localdb` needs pgserver, which has no wheels for Python "
+                f"{sys.version_info.major}.{sys.version_info.minor} (it supports 3.9–3.12).\n"
+                "Either install CodeRAG on Python 3.12:\n"
+                "    pipx install --python python3.12 'coderag-ai[mcp,localdb]'\n"
+                "or run PostgreSQL yourself (Docker: `docker compose up -d db`, or "
+                "Homebrew: `brew install postgresql@16 pgvector`) and set "
+                "CODERAG_DATABASE_URL."
+            ) from exc
         raise RuntimeError(
             "pgserver is not installed. Install the 'localdb' extra:\n"
-            "    pipx install 'coderag[mcp,localdb] @ "
-            "git+https://github.com/galacticsurfer/coderag'"
+            "    pipx install 'coderag-ai[mcp,localdb]'"
         ) from exc
     return pgserver
 
@@ -38,8 +49,27 @@ def start(pgdata: Path | str | None = None) -> str:
     path.parent.mkdir(parents=True, exist_ok=True)
     # cleanup_mode=None -> the postmaster keeps running after we exit.
     server = pgserver.get_server(str(path), cleanup_mode=None)
-    server.psql("CREATE EXTENSION IF NOT EXISTS vector;")
-    return sqlalchemy_url(server.get_uri())
+    url = sqlalchemy_url(server.get_uri())
+    _ensure_vector_extension(url)
+    return url
+
+
+def _ensure_vector_extension(url: str) -> None:
+    """Create the pgvector extension over a normal connection.
+
+    Deliberately does NOT use pgserver's bundled ``psql`` binary: on macOS that
+    executable can fail to run (exit 127), which would break `localdb start` even
+    though the server itself is healthy. psycopg is already a dependency, so we
+    talk to the server the same way the rest of the application does.
+    """
+    from sqlalchemy import create_engine, text
+
+    engine = create_engine(url)
+    try:
+        with engine.begin() as conn:
+            conn.execute(text("CREATE EXTENSION IF NOT EXISTS vector"))
+    finally:
+        engine.dispose()
 
 
 def stop(pgdata: Path | str | None = None, timeout: float = 15.0) -> None:

{coderag_ai-0.3.0 → coderag_ai-0.3.2}/src/coderag/parsing/python.py

@@ -34,9 +34,19 @@ class PythonParser(LanguageParser):
     extensions = (".py",)
 
     def __init__(self) -> None:
-        self._lang = Language(tspython.language(), "python")
-        self._parser = Parser()
-        self._parser.set_language(self._lang)
+        # The tree-sitter Python API changed at 0.22: `Language(ptr)` lost its
+        # `name` argument and `Parser.set_language()` was replaced by passing the
+        # language to the constructor. Support both so the package installs on
+        # any platform/version combination (0.21 has no arm64 macOS wheel).
+        try:
+            self._lang = Language(tspython.language())  # tree-sitter >= 0.22
+        except TypeError:  # pragma: no cover - legacy tree-sitter 0.21
+            self._lang = Language(tspython.language(), "python")  # type: ignore[call-overload]
+        try:
+            self._parser = Parser(self._lang)  # tree-sitter >= 0.22
+        except TypeError:  # pragma: no cover - legacy tree-sitter 0.21
+            self._parser = Parser()
+            self._parser.set_language(self._lang)  # type: ignore[attr-defined]
 
     def parse(self, module_qualified_name: str, source: str) -> ParseResult:
         src = source.encode("utf-8")

coderag_ai-0.3.2/tests/test_phase14_localdb.py

@@ -0,0 +1,72 @@
+"""Rootless local Postgres helper (no Docker/sudo) — lifecycle correctness."""
+
+from __future__ import annotations
+
+import pytest
+
+pytest.importorskip("pgserver")
+
+from coderag import localdb  # noqa: E402
+
+
+def test_status_does_not_start_the_server(tmp_path):
+    """A status check must never start a server (regression: it used to)."""
+    pgdata = tmp_path / "pgdata"
+    assert localdb.status(pgdata) is None       # not running, nothing created
+    assert not (pgdata / "postmaster.pid").exists()
+
+
+def test_start_status_stop_lifecycle(tmp_path):
+    pgdata = tmp_path / "pgdata"
+    url = localdb.start(pgdata)
+    try:
+        assert url.startswith("postgresql+psycopg://")
+        pid = localdb.status(pgdata)
+        assert pid and pid > 0
+    finally:
+        localdb.stop(pgdata)
+    assert localdb.status(pgdata) is None       # stays stopped
+
+
+def test_stale_pidfile_reports_not_running(tmp_path):
+    pgdata = tmp_path / "pgdata"
+    pgdata.mkdir()
+    (pgdata / "postmaster.pid").write_text("999999999\n")  # impossible pid
+    assert localdb.status(pgdata) is None
+
+
+def test_start_does_not_shell_out_to_psql(tmp_path, monkeypatch):
+    """Regression: the bundled `psql` binary can fail on macOS (exit 127).
+
+    `start()` must create the pgvector extension over a normal DB connection, so
+    the command works even when that executable is unusable.
+    """
+    import pgserver
+
+    real_get_server = pgserver.get_server
+
+    def guarded(pgdata, cleanup_mode=None):
+        server = real_get_server(pgdata, cleanup_mode=cleanup_mode)
+
+        def boom(*a, **k):
+            raise AssertionError("start() must not invoke the bundled psql binary")
+
+        monkeypatch.setattr(server, "psql", boom, raising=False)
+        return server
+
+    monkeypatch.setattr(pgserver, "get_server", guarded)
+
+    pgdata = tmp_path / "pgdata"
+    url = localdb.start(pgdata)
+    try:
+        from sqlalchemy import create_engine, text
+
+        engine = create_engine(url)
+        with engine.connect() as conn:
+            got = conn.execute(
+                text("SELECT extversion FROM pg_extension WHERE extname='vector'")
+            ).scalar()
+        engine.dispose()
+        assert got, "pgvector extension should have been created"
+    finally:
+        localdb.stop(pgdata)

coderag_ai-0.3.0/tests/test_phase14_localdb.py

@@ -1,35 +0,0 @@
-"""Rootless local Postgres helper (no Docker/sudo) — lifecycle correctness."""
-
-from __future__ import annotations
-
-import pytest
-
-pytest.importorskip("pgserver")
-
-from coderag import localdb  # noqa: E402
-
-
-def test_status_does_not_start_the_server(tmp_path):
-    """A status check must never start a server (regression: it used to)."""
-    pgdata = tmp_path / "pgdata"
-    assert localdb.status(pgdata) is None       # not running, nothing created
-    assert not (pgdata / "postmaster.pid").exists()
-
-
-def test_start_status_stop_lifecycle(tmp_path):
-    pgdata = tmp_path / "pgdata"
-    url = localdb.start(pgdata)
-    try:
-        assert url.startswith("postgresql+psycopg://")
-        pid = localdb.status(pgdata)
-        assert pid and pid > 0
-    finally:
-        localdb.stop(pgdata)
-    assert localdb.status(pgdata) is None       # stays stopped
-
-
-def test_stale_pidfile_reports_not_running(tmp_path):
-    pgdata = tmp_path / "pgdata"
-    pgdata.mkdir()
-    (pgdata / "postmaster.pid").write_text("999999999\n")  # impossible pid
-    assert localdb.status(pgdata) is None