Merge pull request #32 from Razpines/feat/mcp-source-coverage-search-diagnostics

feat(mcp): add source coverage status and search diagnostics

Author: Razpines

src/unity_docs_mcp/mcp_server.py

@@ -6,7 +6,7 @@
 import os
 import sys
 from pathlib import Path
-from typing import List, Optional
+from typing import Any, List, Optional
 
 from mcp.server.fastmcp import FastMCP
 
@@ -76,17 +76,17 @@ def _response_meta(docstore: DocStore, baked_manifest: Optional[dict] = None) ->
     return meta
 
 
-@app.tool()
-def search(
-    query: str,
-    k: int = 6,
-    source_types: Optional[List[str] | str] = None,
-) -> List[dict]:
-    docstore = _get_docstore()
-    meta = _response_meta(docstore)
+def _parse_source_types(source_types: Optional[List[str] | str]) -> Optional[List[str]]:
+    if source_types is None:
+        return None
     if isinstance(source_types, str):
-        source_types = [s.strip() for s in source_types.split(",") if s.strip()]
-    results = docstore.search(query=query, k=k, source_types=source_types)
+        values = [s.strip().lower() for s in source_types.split(",") if s.strip()]
+    else:
+        values = [str(s).strip().lower() for s in source_types if str(s).strip()]
+    return values or None
+
+
+def _serialize_search_results(results: list[Any], meta: dict) -> List[dict]:
     return [
         {
             "chunk_id": r.chunk_id,
@@ -104,6 +104,65 @@ def search(
     ]
 
 
+@app.tool()
+def search(
+    query: str,
+    k: int = 6,
+    source_types: Optional[List[str] | str] = None,
+    debug: bool = False,
+) -> List[dict] | dict:
+    docstore = _get_docstore()
+    meta = _response_meta(docstore)
+    parsed_source_types = _parse_source_types(source_types)
+    available_source_types = docstore.available_source_types()
+    known_source_types = docstore.known_source_types()
+
+    invalid_source_types = []
+    unavailable_source_types = []
+    if parsed_source_types:
+        requested_set = set(parsed_source_types)
+        known_set = set(known_source_types)
+        available_set = set(available_source_types)
+        invalid_source_types = sorted(requested_set - known_set)
+        unavailable_source_types = sorted((requested_set & known_set) - available_set)
+
+    if invalid_source_types or unavailable_source_types:
+        message_parts = []
+        if invalid_source_types:
+            message_parts.append(f"Unsupported source_types: {', '.join(invalid_source_types)}")
+        if unavailable_source_types:
+            message_parts.append(f"Requested source_types not present in this index: {', '.join(unavailable_source_types)}")
+        return {
+            "error": "invalid_source_types",
+            "message": ". ".join(message_parts),
+            "requested_source_types": parsed_source_types or [],
+            "invalid_source_types": invalid_source_types,
+            "unavailable_source_types": unavailable_source_types,
+            "known_source_types": known_source_types,
+            "available_source_types": available_source_types,
+            "results": [],
+            "meta": meta,
+        }
+
+    results = docstore.search(query=query, k=k, source_types=parsed_source_types)
+    serialized = _serialize_search_results(results, meta)
+    if not debug:
+        return serialized
+    return {
+        "results": serialized,
+        "meta": meta,
+        "debug": {
+            "query": query,
+            "k": k,
+            "requested_source_types": parsed_source_types or [],
+            "available_source_types": available_source_types,
+            "known_source_types": known_source_types,
+            "retrieval_mode": meta["retrieval_mode"],
+            "result_count": len(serialized),
+        },
+    }
+
+
 @app.tool()
 def open(
     doc_id: Optional[str] = None,
@@ -177,13 +236,23 @@ def status() -> dict:
     paths = make_paths(config)
     baked_manifest = _read_manifest(paths.baked_dir / "manifest.json")
     index_manifest = _read_manifest(paths.index_dir / "manifest.json")
+    available_source_types = docstore.available_source_types()
+    source_type_counts = docstore.source_type_counts()
+    coverage_warnings: list[str] = []
+    if "scriptref" not in set(available_source_types):
+        coverage_warnings.append(
+            "scriptref source type is not present in the loaded corpus/index; API symbol lookup coverage may be incomplete"
+        )
     return {
         "meta": _response_meta(docstore, baked_manifest=baked_manifest),
         "paths": vars(config.paths),
         "unity_version": config.unity_version,
         "embedder": vars(config.index.embedder),
         "baked_manifest": baked_manifest,
         "index_manifest": index_manifest,
+        "available_source_types": available_source_types,
+        "source_type_counts": source_type_counts,
+        "coverage_warnings": coverage_warnings,
     }
 
 

src/unity_docs_mcp/tools/ops.py

@@ -10,6 +10,8 @@
 from unity_docs_mcp.index.search import HybridSearcher
 from unity_docs_mcp.paths import make_paths
 
+_DEFAULT_SOURCE_TYPES = ("manual", "scriptref")
+
 
 @dataclass
 class DocRecord:
@@ -26,8 +28,10 @@ def __init__(self, config: Config):
         self.config = config
         self.paths = make_paths(config)
         self.corpus = self._load_corpus(self.paths.baked_dir / "corpus.jsonl")
+        self._doc_source_type_counts = self._count_source_types(self.corpus.values())
         self.link_index = self._load_links(self.paths.baked_dir / "link_graph.jsonl")
         self.searcher = HybridSearcher(config, self.paths.index_dir)
+        self._chunk_source_type_counts = self._count_source_types(getattr(self.searcher, "chunk_meta", {}).values())
 
     def _load_corpus(self, path: Path) -> Dict[str, DocRecord]:
         records: Dict[str, DocRecord] = {}
@@ -56,6 +60,19 @@ def _load_links(self, path: Path) -> Dict[str, List[str]]:
                 links.setdefault(row["from_doc_id"], []).append(row["to_doc_id"])
         return links
 
+    @staticmethod
+    def _count_source_types(rows) -> Dict[str, int]:
+        counts: Dict[str, int] = {}
+        for row in rows:
+            if isinstance(row, dict):
+                source_type = row.get("source_type", "")
+            else:
+                source_type = getattr(row, "source_type", "")
+            if not source_type:
+                continue
+            counts[source_type] = counts.get(source_type, 0) + 1
+        return counts
+
     def open_doc(self, doc_id: Optional[str] = None, path: Optional[str] = None) -> Optional[DocRecord]:
         target_id = doc_id
         if not target_id and path:
@@ -82,3 +99,16 @@ def related(self, doc_id: str, limit: int = 10) -> List[DocRecord]:
 
     def search(self, query: str, k: int = 6, source_types: Optional[List[str]] = None) -> List:
         return self.searcher.search(query=query, k=k, source_types=source_types)
+
+    def available_source_types(self) -> List[str]:
+        all_types = set(self._doc_source_type_counts) | set(self._chunk_source_type_counts)
+        return sorted(all_types)
+
+    def known_source_types(self) -> List[str]:
+        return sorted(set(_DEFAULT_SOURCE_TYPES) | set(self.available_source_types()))
+
+    def source_type_counts(self) -> Dict[str, Dict[str, int]]:
+        return {
+            "docs": dict(sorted(self._doc_source_type_counts.items())),
+            "chunks": dict(sorted(self._chunk_source_type_counts.items())),
+        }

tests/test_mcp_meta.py

@@ -7,6 +7,7 @@
 class _FakeDocStore:
     def __init__(self) -> None:
         self.config = Config()
+        self._available_source_types = ["manual"]
 
     def search(self, query: str, k: int = 6, source_types=None):
         return [
@@ -57,6 +58,15 @@ def related(self, doc_id: str, limit: int = 10):
             )
         ]
 
+    def available_source_types(self):
+        return list(self._available_source_types)
+
+    def known_source_types(self):
+        return ["manual", "scriptref"]
+
+    def source_type_counts(self):
+        return {"docs": {"manual": 1}, "chunks": {"manual": 1}}
+
 
 def _install_fake_docstore(monkeypatch):
     fake = _FakeDocStore()
@@ -91,6 +101,32 @@ def test_open_includes_meta_even_when_missing(monkeypatch):
     assert "doc_id" not in missing
 
 
+def test_search_invalid_source_types_returns_actionable_error(monkeypatch):
+    _install_fake_docstore(monkeypatch)
+    result = mcp_server.search("Rigidbody.AddForce", source_types="scripting")
+    assert result["error"] == "invalid_source_types"
+    assert result["invalid_source_types"] == ["scripting"]
+    assert "manual" in result["available_source_types"]
+
+
+def test_search_unavailable_source_types_returns_actionable_error(monkeypatch):
+    _install_fake_docstore(monkeypatch)
+    result = mcp_server.search("Rigidbody.AddForce", source_types="scriptref")
+    assert result["error"] == "invalid_source_types"
+    assert result["invalid_source_types"] == []
+    assert result["unavailable_source_types"] == ["scriptref"]
+
+
+def test_search_debug_returns_results_and_debug_block(monkeypatch):
+    _install_fake_docstore(monkeypatch)
+    result = mcp_server.search("IJobParallelFor batch size", k=3, debug=True)
+    assert "results" in result
+    assert "debug" in result
+    assert result["debug"]["query"] == "IJobParallelFor batch size"
+    assert result["debug"]["result_count"] == 1
+    assert result["results"][0]["doc_id"] == "manual/job-system-parallel-for-jobs"
+
+
 def test_list_files_and_related_include_meta(monkeypatch):
     _install_fake_docstore(monkeypatch)
     files = mcp_server.list_files("*parallel-for*")
@@ -118,3 +154,6 @@ def _fake_read_manifest(path):
     assert status["meta"]["built_on"] == "2026-02-20"
     assert "paths" in status
     assert "index_manifest" in status
+    assert status["available_source_types"] == ["manual"]
+    assert status["source_type_counts"]["docs"]["manual"] == 1
+    assert status["coverage_warnings"]