local-codebase-indexer 0.3.0__tar.gz → 0.3.2__tar.gz

{local_codebase_indexer-0.3.0/local_codebase_indexer.egg-info → local_codebase_indexer-0.3.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: local-codebase-indexer
-Version: 0.3.0
+Version: 0.3.2
 Summary: Local codebase indexer and context provider for AI coding assistants: MCP server and CLI with SQL data-access and C# call relations. No cloud, no server.
 Author: sfy257000
 Project-URL: Homepage, https://github.com/sfy257000/local-codebase-indexer

{local_codebase_indexer-0.3.0 → local_codebase_indexer-0.3.2}/TODO.md

@@ -4,6 +4,7 @@ Priorities come from the benchmark (`benchmarks/run_benchmark.py`, baseline in `
 
 ## Next
 
+0. **Publish to the MCP registry** (not done yet). PyPI 0.3.0 is out and carries the `mcp-name` marker; `server.json` is ready. Run `mcp-publisher login github` then `mcp-publisher publish` from the repo folder, then verify with `curl "https://registry.modelcontextprotocol.io/v0.1/servers?search=local-codebase-indexer"`. Repeat after each release (bump both versions in `server.json`).
 1. **Check a real embedding model on your own machine** (this environment cannot reach Hugging Face, so the ONNX path is only tested with a tiny generated model). Export a model such as `all-MiniLM-L6-v2` (mean pooling) or `bge-small-en-v1.5` (`INDEXER_ONNX_POOLING=cls`, `INDEXER_ONNX_QUERY_PREFIX="Represent this sentence for searching relevant passages: "`), set `INDEXER_ONNX_MODEL`, and run `python benchmarks/run_benchmark.py --compare benchmarks/baseline.json`.
    Evidence: `synonym` Recall@5 is only 0.30 (hashing embedder plus the dictionary, which cannot cover free paraphrase). Whether a real model fixes this is not measured.
 2. **Resolve `FROM a JOIN b ON ..., c`** (a comma list after a JOIN loses `c`), non-sink SQL proc-name literals (`Run("usp_X")`), and EF query tracing (only the table mapping is recorded now).

{local_codebase_indexer-0.3.0 → local_codebase_indexer-0.3.2}/local_codebase_indexer/__init__.py

@@ -1,3 +1,3 @@
 """Universal Local Codebase Indexer & Context Provider (see index.md)."""
 
-__version__ = "0.3.0"
+__version__ = "0.3.2"

{local_codebase_indexer-0.3.0 → local_codebase_indexer-0.3.2}/local_codebase_indexer/cli.py

@@ -58,8 +58,25 @@ def build_parser() -> argparse.ArgumentParser:
     return p
 
 
+def _fix_stdio(serving: bool) -> None:
+    """MCP speaks UTF-8 JSON; the Windows locale codec (cp950, cp1252) would garble Chinese queries. On a console,
+    keep its encoding but never crash on a character it cannot show."""
+    for name in ("stdin", "stdout", "stderr"):
+        stream = getattr(sys, name, None)
+        if stream is None or not hasattr(stream, "reconfigure"):
+            continue
+        try:
+            if serving:
+                stream.reconfigure(encoding="utf-8", errors="replace")
+            elif name != "stdin":
+                stream.reconfigure(errors="replace")
+        except (OSError, ValueError):
+            pass
+
+
 def main(argv: list[str] | None = None) -> int:
     args = build_parser().parse_args(argv)
+    _fix_stdio(serving=args.cmd == "serve")
     root = os.path.abspath(args.root)
 
     if args.cmd == "serve":

{local_codebase_indexer-0.3.0 → local_codebase_indexer-0.3.2}/local_codebase_indexer/core.py

@@ -61,7 +61,7 @@ def git_info(root: str) -> dict | None:
     """HEAD commit, branch and number of uncommitted paths; None outside a git work tree."""
     def run(*args: str) -> str | None:
         try:
-            r = subprocess.run(["git", "-C", root, *args], capture_output=True, text=True, timeout=15)
+            r = subprocess.run(["git", "-C", root, *args], capture_output=True, text=True, encoding="utf-8", errors="replace", timeout=15)
         except (OSError, subprocess.SubprocessError):
             return None
         return r.stdout if r.returncode == 0 else None

{local_codebase_indexer-0.3.0 → local_codebase_indexer-0.3.2}/local_codebase_indexer/query.py

@@ -94,6 +94,7 @@ class CodebaseTools:
                 terms.append(t)
         ranked: dict[str, list[int]] = {"keyword": [], "symbol": [], "vector": []}
         ranks_override: dict[str, dict[int, int]] = {}
+        exact_ids: set[int] = set()  # chunks whose symbol name contains a full identifier typed in the query
 
         # keyword layer: FTS5 / bm25 over chunk text
         if terms:
@@ -125,6 +126,19 @@ class CodebaseTools:
                 ):
                     if _has_word(t, r["file_path"]):
                         scores[r[0]] += 1
+            # a full identifier typed by the user (SFC_MLot) must match names containing it whole, whatever the
+            # case or underscores (SFC_Mlot, SFC_MLotRepository); sub-word matches alone would favour every *Lot
+            for ident in self._whole_identifiers(query):
+                squashed = re.sub(r"[^a-z0-9]", "", ident.lower())
+                like = f"%{squashed}%"
+                for r in self.conn.execute(
+                    "SELECT c.rowid FROM chunk_metadata c JOIN symbols s ON s.id = c.symbol_id "
+                    "WHERE replace(lower(s.symbol_name), '_', '') LIKE ? LIMIT 500", (like,)):
+                    scores[r[0]] += 3
+                    exact_ids.add(r[0])
+                for r in self.conn.execute(
+                    "SELECT rowid FROM chunk_metadata WHERE replace(lower(file_path), '_', '') LIKE ? LIMIT 300", (like,)):
+                    scores[r[0]] += 2
             ordered = sorted(scores.items(), key=lambda kv: (-kv[1], kv[0]))[:pool]
             ranked["symbol"] = [rid for rid, _ in ordered]
             # equal match counts must share a rank, otherwise arbitrary row order decides
@@ -154,13 +168,26 @@ class CodebaseTools:
                 fused[rid] += LAYER_WEIGHTS[name] / (RRF_K + rank)
                 via[rid].append(name)
         top_ids = [rid for rid, _ in sorted(fused.items(), key=lambda kv: -kv[1])]
+        if exact_ids:  # a typed identifier beats sub-word noise (every *Lot class), ties keep the fused order
+            top_ids = [r for r in top_ids if r in exact_ids] + [r for r in top_ids if r not in exact_ids]
 
         results, seen_spans = [], set()
-        for acc in self._route_data_access(query)[:limit]:
+        routed, bare = self._route_data_access(query)
+        routed_hits = []
+        for acc in routed[:limit]:
             hit = self._chunk_result(acc)
             if hit is not None and (hit["file_path"], hit["start_line"], hit["end_line"]) not in seen_spans:
                 seen_spans.add((hit["file_path"], hit["start_line"], hit["end_line"]))
-                results.append(hit)
+                routed_hits.append(hit)
+        if bare and routed_hits and top_ids:
+            # only a name was typed: the definition (best ordinary match) first, then who touches it
+            first = self._ordinary_hit(top_ids[0], fused, via)
+            if first is not None:
+                span = (first["file_path"], first["start_line"], first["end_line"])
+                routed_hits = [h for h in routed_hits if (h["file_path"], h["start_line"], h["end_line"]) != span]
+                results.append(first)
+                seen_spans.add(span)
+        results.extend(routed_hits)
         for rid in top_ids:
             if len(results) >= limit:
                 break
@@ -194,14 +221,39 @@ class CodebaseTools:
         return {"query": query, "results": results, "related_files": self._related_files(results[:3])}
 
     # ------------------------------------------------------------------ data access routing
+    def _ordinary_hit(self, rid: int, fused: dict, via: dict) -> dict | None:
+        row = self.conn.execute(
+            "SELECT c.file_path, c.start_line, c.end_line, c.raw_text, s.symbol_type, s.qualified_name, s.signature "
+            "FROM chunk_metadata c LEFT JOIN symbols s ON s.id = c.symbol_id WHERE c.rowid=?", (rid,)).fetchone()
+        if row is None:
+            return None
+        lines = textwrap.dedent(row["raw_text"].rstrip()).split("\n")
+        return {"file_path": row["file_path"], "start_line": row["start_line"], "end_line": row["end_line"],
+                "symbol": row["qualified_name"], "symbol_type": row["symbol_type"], "signature": row["signature"],
+                "score": round(fused[rid], 5), "matched_by": via[rid],
+                "snippet": "\n".join(lines[:12]) + ("\n..." if len(lines) > 12 else "")}
+
+    @staticmethod
+    def _whole_identifiers(query: str) -> list[str]:
+        """Identifier-looking tokens (underscore, digit or camelCase inside) of at least 5 characters."""
+        out = []
+        for ident in re.findall(r"[A-Za-z_][\w$]*", query):
+            if len(ident) >= 5 and (re.search(r"[_\d]", ident[1:]) or
+                                    (re.search(r"[a-z]", ident) and re.search(r"[A-Z]", ident[1:]))):
+                out.append(ident)
+        return out
+
     _OBJECT_HINT = re.compile(r"\b(table|tables|procedure|procedures|proc|stored|view|views|sp)\b|資料表|預存程序|存储过程|数据表", re.I)
 
-    def _route_data_access(self, query: str) -> list[dict]:
+    def _route_data_access(self, query: str) -> tuple[list[dict], bool]:
         """Queries such as "what writes to MES_Lot" name a known table / procedure and an intent. A ranked chunk
         list cannot enumerate every writer, so those queries get the accessors from the data access index first."""
         intents = dataflow.detect_intents(query)
+        bare = False
         if not intents:
-            return []
+            if len(query.split()) > 3:
+                return [], False
+            intents, bare = {"touch"}, True
         keys = dataflow.known_object_keys(self.conn)
         hinted = bool(self._OBJECT_HINT.search(query))
         found: list[str] = []
@@ -218,7 +270,9 @@ class CodebaseTools:
                 if sig not in seen:
                     seen.add(sig)
                     out.append(acc)
-        return out
+        if bare:  # leave room for the ordinary name matches (the class, the repository, the file)
+            out = out[:3]
+        return out, bare
 
     def _chunk_result(self, acc: dict) -> dict | None:
         conn = self.conn

{local_codebase_indexer-0.3.0 → local_codebase_indexer-0.3.2/local_codebase_indexer.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: local-codebase-indexer
-Version: 0.3.0
+Version: 0.3.2
 Summary: Local codebase indexer and context provider for AI coding assistants: MCP server and CLI with SQL data-access and C# call relations. No cloud, no server.
 Author: sfy257000
 Project-URL: Homepage, https://github.com/sfy257000/local-codebase-indexer

{local_codebase_indexer-0.3.0 → local_codebase_indexer-0.3.2}/server.json

@@ -7,12 +7,12 @@
     "url": "https://github.com/sfy257000/local-codebase-indexer",
     "source": "github"
   },
-  "version": "0.3.0",
+  "version": "0.3.2",
   "packages": [
     {
       "registryType": "pypi",
       "identifier": "local-codebase-indexer",
-      "version": "0.3.0",
+      "version": "0.3.2",
       "runtimeHint": "uvx",
       "transport": {
         "type": "stdio"

{local_codebase_indexer-0.3.0 → local_codebase_indexer-0.3.2}/tests/test_dataaccess.py

@@ -112,3 +112,33 @@ def test_synonym_expansion_and_project_override(tmp_path):
     e2 = Expander(str(tmp_path))
     assert "wpnlot" in e2.expand("在製批 查詢")
     assert "proc" not in e2.expand("SP list")
+
+
+def test_bare_table_name_finds_its_repository_not_every_lot_class(tmp_path):
+    """Real-world miss: searching SFC_MLot ranked OutsourceLot / PNLToPCSLot above SFC_MLotRepository because the
+    sub-word 'lot' matched everything, and the table was spelled SFC_Mlot in the SQL string."""
+    (tmp_path / "Repo").mkdir()
+    (tmp_path / "Models").mkdir()
+    (tmp_path / "Repo" / "SFC_MLotRepository.cs").write_text("""using System.Threading.Tasks;
+using Dapper;
+namespace E {
+  public class SFC_MLotRepository : ISFC_MLotRepository {
+    private readonly System.Data.IDbConnection _db;
+    public Task<int> Deduct(decimal qty) {
+        string sql = @" UPDATE SFC_Mlot SET Qty = Qty - @qty WHERE PostFlag = 'W'; DELETE FROM SFC_Mlot WHERE Qty <= 0;";
+        return _db.ExecuteAsync(sql, new { qty });
+    }
+  }
+  public interface ISFC_MLotRepository { Task<int> Deduct(decimal qty); }
+}
+""", encoding="utf-8")
+    lots = "".join(f"public class Lot{i} : SFCLot {{ }} public class ReworkLot{i} : SFCLot {{ }}\n" for i in range(8))
+    (tmp_path / "Models" / "SFCLot.cs").write_text(
+        "namespace E { public partial class SFCLot { }\n" + lots + "public class OutsourceLot : SFCLot { } }", encoding="utf-8")
+    t = CodebaseTools(str(tmp_path))
+    for q in ("SFC_MLot", "SFC_Mlot", "sfc_mlot"):
+        top = t.semantic_search(q, limit=5)["results"]
+        assert top[0]["file_path"].endswith("SFC_MLotRepository.cs"), q
+        assert sum(r["file_path"].endswith("SFC_MLotRepository.cs") for r in top) >= 3, q
+        assert any((r.get("data_access") or {}).get("access") == "write" for r in top), q
+    t.indexer.close()

{local_codebase_indexer-0.3.0 → local_codebase_indexer-0.3.2}/tests/test_indexer.py

@@ -330,3 +330,30 @@ def test_benchmark_corpus_is_valid_and_floor_holds(tmp_path):
     cats = json.loads(out.read_text(encoding="utf-8"))["suites"]["synthetic"]["categories"]
     assert cats["exact_identifier"]["recall5"] == 1.0
     assert cats["chinese"]["recall5"] == 1.0
+
+
+def test_git_output_is_decoded_as_utf8_not_the_locale(tmp_path, monkeypatch):
+    """On Windows the locale codec (cp950) cannot decode git's UTF-8 output and the reader thread crashed."""
+    import subprocess
+
+    import local_codebase_indexer.core as core
+    seen = {}
+    real = subprocess.run
+
+    def spy(*a, **kw):
+        seen.update(kw)
+        return real(*a, **kw)
+
+    monkeypatch.setattr(subprocess, "run", spy)
+    core.git_info(str(tmp_path))
+    assert seen.get("encoding") == "utf-8" and seen.get("errors") == "replace"
+
+
+def test_mcp_serve_reads_utf8_even_when_locale_is_cp950(project):
+    q = json.dumps({"jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": {
+        "name": "codebase_semantic_search", "arguments": {"query": "批號 LotNo", "limit": 2}}}, ensure_ascii=False)
+    env = dict(os.environ, PYTHONIOENCODING="cp950")
+    r = subprocess.run([sys.executable, "-m", "local_codebase_indexer", "--root", str(project), "serve"],
+                       input=(q + "\n").encode("utf-8"), capture_output=True, env=env, cwd=ROOT, timeout=60)
+    out = json.loads(r.stdout.decode("utf-8"))
+    assert "result" in out and "error" not in out, r.stderr.decode("utf-8", "replace")