Buckets:

download
raw
35.7 kB
import time from pathlib import Path from types import SimpleNamespace from typing import Any, cast import pytest from mcp_server.indexer.bm25_indexer import BM25Indexer from mcp_server.indexer.hybrid_search import HybridSearch, HybridSearchConfig from mcp_server.storage.sqlite_store import SQLiteStore from mcp_server.utils import semantic_indexer as semantic_indexer_module from mcp_server.utils.semantic_indexer import SemanticIndexer from scripts.run_e2e_retrieval_validation import _collect_files class _SemanticStub: def __init__(self) -> None: self.is_available = True self.connection_mode = "test" def validate_connection(self) -> bool: return True def search(self, query: str, limit: int = 20): return [ { "relative_path": "mcp_server/cli/setup_commands.py", "score": 0.91, "content": f"match for {query}", } ][:limit] class _FuzzyStub: def search(self, query: str, limit: int = 20): return [ { "file": "mcp_server/setup/semantic_preflight.py", "snippet": f"fuzzy hit {query}", } ][:limit] class _BM25Stub: def search(self, query: str, limit: int = 20, **kwargs): return [ { "filepath": "mcp_server/cli/setup_commands.py", "score": 42.0, "snippet": f"bm25 hit {query}", } ][:limit] class _SlowSemanticStub(_SemanticStub): def search(self, query: str, limit: int = 20): time.sleep(0.2) return super().search(query, limit) def test_classic_search_code_fts_handles_string_file_ids(temp_db_path): store = SQLiteStore(str(temp_db_path)) repo_id = store.create_repository("/repo", "repo") store.store_file( repository_id=repo_id, path="/repo/mcp_server/setup/semantic_preflight.py", relative_path="mcp_server/setup/semantic_preflight.py", language="python", ) with store._get_connection() as conn: conn.execute( "INSERT INTO fts_code (content, file_id) VALUES (?, ?)", ( "semantic preflight readiness checks", "mcp_server/setup/semantic_preflight.py", ), ) results = store.search_code_fts("semantic preflight", limit=5) assert results assert "semantic_preflight.py" in str(results[0].get("file_path", "")) def test_classic_search_code_fts_demotes_init_for_filename_match(temp_db_path): store = SQLiteStore(str(temp_db_path)) repo_id = store.create_repository("/repo", "repo") store.store_file( repository_id=repo_id, path="/repo/mcp_server/setup/__init__.py", relative_path="mcp_server/setup/__init__.py", language="python", ) store.store_file( repository_id=repo_id, path="/repo/mcp_server/setup/semantic_preflight.py", relative_path="mcp_server/setup/semantic_preflight.py", language="python", ) with store._get_connection() as conn: conn.execute( "INSERT INTO fts_code (content, file_id) VALUES (?, ?)", ( "semantic preflight exports and helpers", "mcp_server/setup/__init__.py", ), ) conn.execute( "INSERT INTO fts_code (content, file_id) VALUES (?, ?)", ( "semantic preflight readiness checks and validation", "mcp_server/setup/semantic_preflight.py", ), ) results = store.search_code_fts("semantic preflight", limit=5) assert results assert results[0]["file_path"].endswith("semantic_preflight.py") def test_bm25_search_demotes_init_for_symbol_query(temp_db_path): store = SQLiteStore(str(temp_db_path)) repo_id = store.create_repository("/repo", "repo") init_file_id = store.store_file( repository_id=repo_id, path="/repo/mcp_server/utils/__init__.py", relative_path="mcp_server/utils/__init__.py", language="python", ) impl_file_id = store.store_file( repository_id=repo_id, path="/repo/mcp_server/utils/semantic_indexer.py", relative_path="mcp_server/utils/semantic_indexer.py", language="python", ) store.store_symbol( file_id=init_file_id, name="SemanticIndexer", kind="class", line_start=1, line_end=1, signature="from .semantic_indexer import SemanticIndexer", documentation="", ) store.store_symbol( file_id=impl_file_id, name="SemanticIndexer", kind="class", line_start=1, line_end=2, signature="class SemanticIndexer:", documentation="", ) bm25 = BM25Indexer(storage=store) bm25.add_document( "mcp_server/utils/__init__.py", "from .semantic_indexer import SemanticIndexer", {"language": "python"}, ) bm25.add_document( "mcp_server/utils/semantic_indexer.py", "class SemanticIndexer:\n pass", {"language": "python"}, ) results = bm25.search("class SemanticIndexer", limit=5) assert results assert str(results[0]["filepath"]).endswith("semantic_indexer.py") @pytest.mark.asyncio async def test_hybrid_handles_sparse_branch_payloads(temp_db_path): store = SQLiteStore(str(temp_db_path)) hybrid = HybridSearch( storage=store, bm25_indexer=None, semantic_indexer=cast(Any, _SemanticStub()), fuzzy_indexer=_FuzzyStub(), config=HybridSearchConfig( enable_bm25=False, enable_semantic=True, enable_fuzzy=True, cache_results=False, parallel_execution=True, individual_limit=5, final_limit=5, ), ) results = await hybrid.search( "how does semantic setup validate qdrant and embedding readiness", limit=5, ) assert results assert results[0].get("filepath") @pytest.mark.asyncio async def test_hybrid_parallel_timeout_preserves_partial_results(temp_db_path): store = SQLiteStore(str(temp_db_path)) hybrid = HybridSearch( storage=store, bm25_indexer=cast(Any, _BM25Stub()), semantic_indexer=cast(Any, _SlowSemanticStub()), fuzzy_indexer=None, config=HybridSearchConfig( enable_bm25=True, enable_semantic=True, enable_fuzzy=False, cache_results=False, parallel_execution=True, parallel_timeout_seconds=0.05, individual_limit=5, final_limit=5, ), ) results = await hybrid.search("semantic setup", limit=5) assert results assert any("setup_commands.py" in str(r.get("filepath", "")) for r in results) def test_rrf_prefers_best_rank_for_result_metadata(temp_db_path): store = SQLiteStore(str(temp_db_path)) hybrid = HybridSearch(storage=store, config=HybridSearchConfig()) from mcp_server.indexer.hybrid_search import SearchResult result_lists = [ [ SearchResult( doc_id="doc-1", filepath="mcp_server/cli/setup_commands.py", score=0.1, snippet="rank-1", metadata={"origin": "bm25"}, source="bm25", ) ], [ SearchResult( doc_id="doc-1", filepath="README.md", score=999.0, snippet="rank-5", metadata={"origin": "semantic"}, source="semantic", ) ], ] fused = hybrid._reciprocal_rank_fusion(result_lists) assert fused assert fused[0].filepath == "mcp_server/cli/setup_commands.py" def test_index_file_uses_chunker_retrieval_metadata(monkeypatch, tmp_path): file_path = tmp_path / "example.py" file_path.write_text("def index_file(path):\n return path\n", encoding="utf-8") fake_chunk = SimpleNamespace( content="def index_file(path):\n return path", metadata={ "symbol": "index_file", "kind": "function", "signature_text": "index_file(path)", "parent_symbol": "SemanticIndexer", "qualified_name": "SemanticIndexer.index_file", "semantic_path": "example.py::SemanticIndexer.index_file", "semantic_text": "function index_file path SemanticIndexer TreeSitterWrapper", "imports": ["from .treesitter_wrapper import TreeSitterWrapper"], }, chunk_id="chunk-1", node_id="node-1", start_line=1, end_line=2, node_type="function_definition", ) captured = {} class _QdrantStub: def upsert(self, collection_name, points): captured["collection_name"] = collection_name captured["points"] = points monkeypatch.setattr(semantic_indexer_module, "chunk_file", lambda *args, **kwargs: [fake_chunk]) indexer = object.__new__(SemanticIndexer) indexer.path_resolver = cast(Any, SimpleNamespace(normalize_path=lambda _: "tmp/example.py")) indexer.collection = "test-collection" indexer._qdrant_available = True indexer.qdrant = cast(Any, _QdrantStub()) embed_calls = [] def _embed_texts(texts, input_type="document"): embed_calls.extend(texts) return [[0.1, 0.2, 0.3] for _ in texts] indexer._embed_texts = _embed_texts result = SemanticIndexer.index_file(indexer, Path(file_path)) assert result["symbols"][0]["symbol"] == "index_file" assert result["chunk_count"] == 1 assert result["embedding_unit_count"] == 1 assert result["file_summary_indexed"] is True assert result["used_fallback_chunks"] is False point = captured["points"][0] payload = point.payload assert payload["symbol"] == "index_file" assert payload["kind"] == "function" assert payload["signature"] == "index_file(path)" assert payload["qualified_name"] == "SemanticIndexer.index_file" assert payload["semantic_text"].startswith("function index_file") assert "file tmp/example.py" in payload["embedding_text"] assert "module example" in payload["embedding_text"] assert "qualified name SemanticIndexer.index_file" in payload["embedding_text"] assert "imports from .treesitter_wrapper import TreeSitterWrapper" in payload["embedding_text"] assert embed_calls[0] == payload["embedding_text"] assert len(embed_calls) == 2 assert "symbols index_file" in embed_calls[1] assert captured["points"][-1].payload["chunk_id"] == "tmp/example.py:file-summary" def test_index_file_splits_oversize_embedding_units(monkeypatch, tmp_path): file_path = tmp_path / "long_example.py" file_path.write_text("def index_file(path):\n return path\n", encoding="utf-8") long_content = "\n\n".join([f"block_{idx}\n" + ("x" * 320) for idx in range(1, 8)]) fake_chunk = SimpleNamespace( content=long_content, metadata={ "symbol": "index_file", "kind": "function", "signature_text": "index_file(path)", "semantic_text": long_content, }, chunk_id="chunk-1", node_id="node-1", start_line=1, end_line=2, node_type="function_definition", ) monkeypatch.setenv("SEMANTIC_MAX_EMBED_CHARS", "1200") monkeypatch.setattr(semantic_indexer_module, "chunk_file", lambda *args, **kwargs: [fake_chunk]) captured = {} class _QdrantStub: def upsert(self, collection_name, points): captured["points"] = points indexer = object.__new__(SemanticIndexer) indexer.path_resolver = cast( Any, SimpleNamespace(normalize_path=lambda _: "tmp/long_example.py") ) indexer.collection = "test-collection" indexer._qdrant_available = True indexer.qdrant = cast(Any, _QdrantStub()) embedded = [] def _embed_texts(texts, input_type="document"): embedded.extend(texts) return [[0.1, 0.2, 0.3] for _ in texts] indexer._embed_texts = _embed_texts SemanticIndexer.index_file(indexer, Path(file_path)) payloads = [point.payload for point in captured["points"]] chunk_payloads = [payload for payload in payloads if payload["kind"] != "file_summary"] assert len(chunk_payloads) > 1 assert all(len(payload["embedding_text"]) <= 1200 for payload in chunk_payloads) assert [payload["subchunk_index"] for payload in chunk_payloads] == list( range(1, len(chunk_payloads) + 1) ) assert all(payload["subchunk_total"] == len(chunk_payloads) for payload in chunk_payloads) assert all(payload["source_chunk_id"] == "chunk-1" for payload in chunk_payloads) assert any("chunk part 1 of" in text for text in embedded) assert payloads[-1]["kind"] == "file_summary" def test_build_chunk_embedding_text_adds_symbol_extraction_summary(monkeypatch): monkeypatch.setenv("SEMANTIC_MAX_EMBED_CHARS", "8000") indexer = object.__new__(SemanticIndexer) embedding_text = SemanticIndexer._build_chunk_embedding_text( indexer, relative_path="mcp_server/utils/semantic_indexer.py", symbol_name="index_file", kind="function", signature="def index_file(self, path: Path) -> dict[str, Any]:", parent_symbol=None, metadata={}, chunk_content=( "chunk_results = chunk_file(path, 'python', extract_metadata=True)\n" "symbols.append({'symbol': symbol_name})" ), ) assert "extract symbols from python files" in embedding_text assert "semantic indexer extracts symbols from python using treesitter" in embedding_text def test_build_file_embedding_text_adds_semantic_indexer_summary(monkeypatch): monkeypatch.setenv("SEMANTIC_MAX_EMBED_CHARS", "8000") indexer = object.__new__(SemanticIndexer) embedding_text = SemanticIndexer._build_file_embedding_text( indexer, relative_path="mcp_server/utils/semantic_indexer.py", symbols=[{"symbol": "index_file", "kind": "function"}], normalized_chunks=[ { "embedding_text": ( "chunk_results = chunk_file(path, 'python', extract_metadata=True)\n" "symbols.append({'symbol': symbol_name})" ) } ], ) assert "symbols index_file" in embedding_text assert "semantic indexer extracts symbols from python using treesitter" in embedding_text def test_index_file_falls_back_to_text_chunks_for_unknown_language(monkeypatch, tmp_path): file_path = tmp_path / "NOTES.md" file_path.write_text("# Notes\n\nFirst paragraph.\n\nSecond paragraph.", encoding="utf-8") def _raise_chunk_error(*args, **kwargs): raise RuntimeError("no grammar available") monkeypatch.setattr(semantic_indexer_module, "chunk_file", _raise_chunk_error) captured = {} class _QdrantStub: def upsert(self, collection_name, points): captured["points"] = points indexer = object.__new__(SemanticIndexer) indexer.path_resolver = cast(Any, SimpleNamespace(normalize_path=lambda _: "tmp/NOTES.md")) indexer.collection = "test-collection" indexer._qdrant_available = True indexer.qdrant = cast(Any, _QdrantStub()) indexer._embed_texts = lambda texts, input_type="document": [[0.1, 0.2, 0.3] for _ in texts] result = SemanticIndexer.index_file(indexer, Path(file_path)) assert result["language"] == "markdown" assert result["used_fallback_chunks"] is True payloads = [point.payload for point in captured["points"]] assert payloads[0]["language"] == "markdown" assert payloads[0]["kind"] == "text" assert "First paragraph." in payloads[0]["embedding_text"] def test_semantic_query_reranks_code_paths_for_code_intent(): indexer = object.__new__(SemanticIndexer) results = [ { "relative_path": "ai_docs/tree_sitter_overview.md", "doc_type": "markdown", "semantic_text": "extract symbols from python using treesitter overview", "score": 0.92, }, { "relative_path": "mcp_server/utils/semantic_indexer.py", "language": "python", "semantic_text": "function index_file extracts symbols from python using treesitter metadata", "qualified_name": "SemanticIndexer.index_file", "score": 0.84, }, ] reranked = SemanticIndexer._rerank_query_results( indexer, "extract symbols from python using treesitter", results, limit=2, ) assert reranked[0]["relative_path"] == "mcp_server/utils/semantic_indexer.py" def test_semantic_query_prefers_semantic_indexer_over_generic_treesitter_plugin(): indexer = object.__new__(SemanticIndexer) results = [ { "relative_path": "mcp_server/plugins/generic_treesitter_plugin.py", "semantic_text": "generic treesitter plugin extracts symbols from source files", "embedding_text": "tree sitter generic plugin parser symbols", "score": 0.9, }, { "relative_path": "mcp_server/utils/semantic_indexer.py", "semantic_text": "semantic indexer extracts symbols from python using treesitter", "embedding_text": "file semantic_indexer.py module semantic indexer extract symbols index embeddings treesitter", "score": 0.82, }, ] reranked = SemanticIndexer._rerank_query_results( indexer, "extract symbols from python using treesitter", results, limit=2, ) assert reranked[0]["relative_path"] == "mcp_server/utils/semantic_indexer.py" def test_semantic_query_deduplicates_chunks_per_file(): indexer = object.__new__(SemanticIndexer) results = [ { "relative_path": "mcp_server/plugins/generic_treesitter_plugin.py", "semantic_text": "generic treesitter plugin extracts symbols from source files", "score": 0.9, }, { "relative_path": "mcp_server/plugins/generic_treesitter_plugin.py", "semantic_text": "generic treesitter plugin parser chunks and symbols", "score": 0.89, }, { "relative_path": "mcp_server/utils/semantic_indexer.py", "semantic_text": "semantic indexer extracts symbols from python using treesitter", "score": 0.82, }, ] reranked = SemanticIndexer._rerank_query_results( indexer, "extract symbols from python using treesitter", results, limit=3, ) assert [item["relative_path"] for item in reranked] == [ "mcp_server/utils/semantic_indexer.py", "mcp_server/plugins/generic_treesitter_plugin.py", ] def test_semantic_query_overlap_uses_relative_path_tokens(): indexer = object.__new__(SemanticIndexer) results = [ { "relative_path": "mcp_server/watcher_multi_repo.py", "semantic_text": "artifact sync watcher", "score": 0.9, }, { "relative_path": "mcp_server/artifacts/delta_resolver.py", "semantic_text": "resolve delta chain for target commit", "score": 0.82, }, ] reranked = SemanticIndexer._rerank_query_results( indexer, "how do artifact push pull and delta resolution work", results, limit=2, ) assert reranked[0]["relative_path"] == "mcp_server/artifacts/delta_resolver.py" def test_semantic_query_symbol_precise_demotes_tests(): indexer = object.__new__(SemanticIndexer) results = [ { "relative_path": "tests/test_profile_aware_semantic_indexer.py", "symbol": "SemanticIndexer", "semantic_text": "test SemanticIndexer behavior", "score": 0.92, }, { "relative_path": "mcp_server/utils/semantic_indexer.py", "symbol": "SemanticIndexer", "qualified_name": "SemanticIndexer", "semantic_text": "SemanticIndexer implementation", "score": 0.84, }, ] reranked = SemanticIndexer._rerank_query_results( indexer, "class SemanticIndexer", results, limit=2, ) assert reranked[0]["relative_path"] == "mcp_server/utils/semantic_indexer.py" def test_semantic_query_symbol_precise_prefers_impl_over_plans_docs_and_benchmarks(): indexer = object.__new__(SemanticIndexer) results = [ { "relative_path": "plans/phase-plan-v7-SEMQUERY.md", "symbol": "SemanticIndexer", "semantic_text": "phase plan for SemanticIndexer query routing", "score": 0.98, }, { "relative_path": "docs/guides/semantic-onboarding.md", "symbol": "SemanticIndexer", "doc_type": "markdown", "semantic_text": "SemanticIndexer documentation", "score": 0.96, }, { "relative_path": "docs/benchmarks/semantic-search-report.md", "symbol": "SemanticIndexer", "doc_type": "markdown", "semantic_text": "benchmark fixture mentions SemanticIndexer", "score": 0.95, }, { "relative_path": "tests/test_profile_aware_semantic_indexer.py", "symbol": "SemanticIndexer", "semantic_text": "test SemanticIndexer behavior", "score": 0.94, }, { "relative_path": "mcp_server/utils/semantic_indexer.py", "symbol": "SemanticIndexer", "qualified_name": "SemanticIndexer", "semantic_text": "SemanticIndexer implementation", "score": 0.84, }, ] reranked = SemanticIndexer._rerank_query_results( indexer, "class SemanticIndexer", results, limit=5, ) assert reranked[0]["relative_path"] == "mcp_server/utils/semantic_indexer.py" assert reranked[-1]["relative_path"] == "plans/phase-plan-v7-SEMQUERY.md" @pytest.mark.parametrize( ("query", "results", "expected_top"), [ ( "how does semantic setup validate qdrant and embedding readiness", [ { "relative_path": "docs/guides/semantic-onboarding.md", "doc_type": "markdown", "semantic_text": "guide for semantic setup and readiness", "score": 0.94, }, { "relative_path": "mcp_server/setup/semantic_preflight.py", "semantic_text": "validate qdrant collection and embedding dimension readiness", "score": 0.83, }, { "relative_path": "mcp_server/cli/setup_commands.py", "semantic_text": "setup semantic CLI command runs semantic preflight", "score": 0.81, }, ], "mcp_server/cli/setup_commands.py", ), ( "where does repository status print semantic readiness evidence", [ { "relative_path": "docs/status/SEMANTIC_DOGFOOD_REBUILD.md", "doc_type": "markdown", "semantic_text": "dogfood report for semantic readiness evidence", "score": 0.95, }, { "relative_path": "mcp_server/cli/repository_commands.py", "semantic_text": "repository status command prints semantic readiness evidence", "score": 0.82, }, { "relative_path": "tests/test_repository_commands.py", "semantic_text": "repository status CLI tests", "score": 0.8, }, ], "mcp_server/cli/repository_commands.py", ), ], ) def test_semdogfood_queries_prefer_repo_implementation_paths(query, results, expected_top): indexer = object.__new__(SemanticIndexer) reranked = SemanticIndexer._rerank_query_results(indexer, query, results, limit=3) assert reranked[0]["relative_path"] == expected_top def test_semdogfood_symbol_query_prefers_repository_status_implementation_over_docs(): indexer = object.__new__(SemanticIndexer) results = [ { "relative_path": "docs/status/SEMANTIC_DOGFOOD_REBUILD.md", "symbol": "status", "doc_type": "markdown", "semantic_text": "dogfood report cites repository status output", "score": 0.96, }, { "relative_path": "mcp_server/cli/repository_commands.py", "symbol": "status", "qualified_name": "status", "semantic_text": "status command prints lexical and semantic readiness", "score": 0.82, }, ] reranked = SemanticIndexer._rerank_query_results(indexer, "function status", results, limit=2) assert reranked[0]["relative_path"] == "mcp_server/cli/repository_commands.py" def test_semantic_query_results_include_stable_metadata(): indexer = object.__new__(SemanticIndexer) indexer._qdrant_available = True indexer.collection = "code_index__oss_high__v1" indexer.semantic_profile = cast(Any, SimpleNamespace(profile_id="oss_high")) indexer._embed_texts = lambda texts, input_type="query": [[0.1, 0.2, 0.3] for _ in texts] class _Point: def __init__(self): self.payload = { "relative_path": "mcp_server/utils/semantic_indexer.py", "line": 42, "semantic_text": "SemanticIndexer implementation", } self.score = 0.91 class _QdrantStub: def search(self, collection_name, query_vector, limit): return [_Point()] indexer.qdrant = cast(Any, _QdrantStub()) results = list(SemanticIndexer.query(indexer, "class SemanticIndexer", limit=1)) assert results[0]["source"] == "semantic" assert results[0]["semantic_source"] == "semantic" assert results[0]["semantic_profile_id"] == "oss_high" assert results[0]["semantic_collection_name"] == "code_index__oss_high__v1" def test_hybrid_post_process_prefers_impl_paths_for_code_intent(temp_db_path): store = SQLiteStore(str(temp_db_path)) hybrid = HybridSearch(storage=store, config=HybridSearchConfig()) from mcp_server.indexer.hybrid_search import SearchResult results = [ SearchResult( doc_id="docs", filepath="ai_docs/tree_sitter_overview.md", score=0.9, snippet="docs", metadata={}, source="semantic", ), SearchResult( doc_id="code", filepath="mcp_server/utils/semantic_indexer.py", score=0.8, snippet="code", metadata={}, source="semantic", ), ] reranked = hybrid._post_process_results( results, limit=2, query="extract symbols from python using treesitter", ) assert reranked[0].filepath == "mcp_server/utils/semantic_indexer.py" def test_hybrid_post_process_uses_filepath_overlap(temp_db_path): store = SQLiteStore(str(temp_db_path)) hybrid = HybridSearch(storage=store, config=HybridSearchConfig()) from mcp_server.indexer.hybrid_search import SearchResult results = [ SearchResult( doc_id="watcher", filepath="mcp_server/watcher_multi_repo.py", score=0.9, snippet="artifact sync watcher", metadata={}, source="semantic", ), SearchResult( doc_id="delta", filepath="mcp_server/artifacts/delta_resolver.py", score=0.8, snippet="resolve delta chain for target commit", metadata={}, source="semantic", ), ] reranked = hybrid._post_process_results( results, limit=2, query="how do artifact push pull and delta resolution work", ) assert reranked[0].filepath == "mcp_server/artifacts/delta_resolver.py" def test_hybrid_post_process_prefers_setup_commands_for_setup_validation(temp_db_path): store = SQLiteStore(str(temp_db_path)) hybrid = HybridSearch(storage=store, config=HybridSearchConfig()) from mcp_server.indexer.hybrid_search import SearchResult results = [ SearchResult( doc_id="preflight", filepath="mcp_server/setup/semantic_preflight.py", score=0.92, snippet="validate qdrant and embedding readiness preflight", metadata={}, source="semantic", ), SearchResult( doc_id="setup", filepath="mcp_server/cli/setup_commands.py", score=0.83, snippet="setup command validates qdrant and embedding readiness", metadata={}, source="bm25", ), ] reranked = hybrid._post_process_results( results, limit=2, query="how does semantic setup validate qdrant and embedding readiness", ) assert reranked[0].filepath == "mcp_server/cli/setup_commands.py" def test_hybrid_post_process_prefers_delta_resolver_for_delta_resolution(temp_db_path): store = SQLiteStore(str(temp_db_path)) hybrid = HybridSearch(storage=store, config=HybridSearchConfig()) from mcp_server.indexer.hybrid_search import SearchResult results = [ SearchResult( doc_id="artifacts", filepath="mcp_server/artifacts/delta_artifacts.py", score=0.9, snippet="artifact push pull state", metadata={}, source="semantic", ), SearchResult( doc_id="resolver", filepath="mcp_server/artifacts/delta_resolver.py", score=0.82, snippet="resolve delta chain for artifact pull", metadata={}, source="bm25", ), ] reranked = hybrid._post_process_results( results, limit=2, query="how do artifact push pull and delta resolution work", ) assert reranked[0].filepath == "mcp_server/artifacts/delta_resolver.py" def test_hybrid_post_process_prefers_delta_resolver_over_delta_artifacts(temp_db_path): store = SQLiteStore(str(temp_db_path)) hybrid = HybridSearch(storage=store, config=HybridSearchConfig()) from mcp_server.indexer.hybrid_search import SearchResult results = [ SearchResult( doc_id="artifacts", filepath="mcp_server/artifacts/delta_artifacts.py", score=0.96, snippet=( "build delta archive manifest operations checksums apply artifact " "push pull workflow" ), metadata={}, source="semantic", ), SearchResult( doc_id="resolver", filepath="mcp_server/artifacts/delta_resolver.py", score=0.82, snippet=( "resolve delta chain base_commit target_commit for artifact pull " "resolution" ), metadata={}, source="bm25", ), ] reranked = hybrid._post_process_results( results, limit=2, query="how do artifact push pull and delta resolution work", ) assert reranked[0].filepath == "mcp_server/artifacts/delta_resolver.py" def test_hybrid_post_process_demotes_artifact_commands_for_resolution_query( temp_db_path, ): store = SQLiteStore(str(temp_db_path)) hybrid = HybridSearch(storage=store, config=HybridSearchConfig()) from mcp_server.indexer.hybrid_search import SearchResult results = [ SearchResult( doc_id="commands", filepath="mcp_server/cli/artifact_commands.py", score=0.92, snippet="artifact push and pull CLI commands", metadata={}, source="bm25", ), SearchResult( doc_id="resolver", filepath="mcp_server/artifacts/delta_resolver.py", score=0.81, snippet="resolve delta chain for artifact pull target commit", metadata={}, source="semantic", ), ] reranked = hybrid._post_process_results( results, limit=2, query="how do artifact push pull and delta resolution work", ) assert reranked[0].filepath == "mcp_server/artifacts/delta_resolver.py" def test_hybrid_post_process_symbol_precise_demotes_tests(temp_db_path): store = SQLiteStore(str(temp_db_path)) hybrid = HybridSearch(storage=store, config=HybridSearchConfig()) from mcp_server.indexer.hybrid_search import SearchResult results = [ SearchResult( doc_id="test", filepath="tests/test_profile_aware_semantic_indexer.py", score=0.9, snippet="class SemanticIndexer test", metadata={}, source="semantic", ), SearchResult( doc_id="impl", filepath="mcp_server/utils/semantic_indexer.py", score=0.82, snippet="class SemanticIndexer implementation", metadata={}, source="semantic", ), ] reranked = hybrid._post_process_results( results, limit=2, query="class SemanticIndexer", ) assert reranked[0].filepath == "mcp_server/utils/semantic_indexer.py" def test_collect_files_excludes_generated_dirs_and_prioritizes_source(tmp_path): (tmp_path / "mcp_server").mkdir() (tmp_path / "mcp_server" / "core.py").write_text("print('core')\n", encoding="utf-8") (tmp_path / "scripts").mkdir() (tmp_path / "scripts" / "runner.py").write_text("print('runner')\n", encoding="utf-8") (tmp_path / "docs").mkdir() (tmp_path / "docs" / "guide.md").write_text("# Guide\n", encoding="utf-8") (tmp_path / "docs" / "benchmarks").mkdir(parents=True) (tmp_path / "docs" / "benchmarks" / "report.md").write_text("# Report\n", encoding="utf-8") (tmp_path / "analysis_archive").mkdir() (tmp_path / "analysis_archive" / "notes.md").write_text("# Notes\n", encoding="utf-8") (tmp_path / "tests").mkdir() (tmp_path / "tests" / "test_core.py").write_text( "def test_core():\n pass\n", encoding="utf-8" ) (tmp_path / "htmlcov").mkdir() (tmp_path / "htmlcov" / "status.json").write_text("{}\n", encoding="utf-8") files = _collect_files(tmp_path, max_files=10) relative_paths = [path.relative_to(tmp_path).as_posix() for path in files] assert "mcp_server/core.py" in relative_paths assert "scripts/runner.py" in relative_paths assert "docs/guide.md" in relative_paths assert "tests/test_core.py" in relative_paths assert "docs/benchmarks/report.md" not in relative_paths assert "analysis_archive/notes.md" not in relative_paths assert "htmlcov/status.json" not in relative_paths assert relative_paths.index("mcp_server/core.py") < relative_paths.index("docs/guide.md") assert relative_paths.index("docs/guide.md") < relative_paths.index("tests/test_core.py")
import time
from pathlib import Path
from types import SimpleNamespace
from typing import Any, cast
import pytest
from mcp_server.indexer.bm25_indexer import BM25Indexer
from mcp_server.indexer.hybrid_search import HybridSearch, HybridSearchConfig
from mcp_server.storage.sqlite_store import SQLiteStore
from mcp_server.utils import semantic_indexer as semantic_indexer_module
from mcp_server.utils.semantic_indexer import SemanticIndexer
from scripts.run_e2e_retrieval_validation import _collect_files
class _SemanticStub:
def __init__(self) -> None:
self.is_available = True
self.connection_mode = "test"
def validate_connection(self) -> bool:
return True
def search(self, query: str, limit: int = 20):
return [
{
"relative_path": "mcp_server/cli/setup_commands.py",
"score": 0.91,
"content": f"match for {query}",
}
][:limit]
class _FuzzyStub:
def search(self, query: str, limit: int = 20):
return [
{
"file": "mcp_server/setup/semantic_preflight.py",
"snippet": f"fuzzy hit {query}",
}
][:limit]
class _BM25Stub:
def search(self, query: str, limit: int = 20, **kwargs):
return [
{
"filepath": "mcp_server/cli/setup_commands.py",
"score": 42.0,
"snippet": f"bm25 hit {query}",
}
][:limit]
class _SlowSemanticStub(_SemanticStub):
def search(self, query: str, limit: int = 20):
time.sleep(0.2)
return super().search(query, limit)
def test_classic_search_code_fts_handles_string_file_ids(temp_db_path):
store = SQLiteStore(str(temp_db_path))
repo_id = store.create_repository("/repo", "repo")
store.store_file(
repository_id=repo_id,
path="/repo/mcp_server/setup/semantic_preflight.py",
relative_path="mcp_server/setup/semantic_preflight.py",
language="python",
)
with store._get_connection() as conn:
conn.execute(
"INSERT INTO fts_code (content, file_id) VALUES (?, ?)",
(
"semantic preflight readiness checks",
"mcp_server/setup/semantic_preflight.py",
),
)
results = store.search_code_fts("semantic preflight", limit=5)
assert results
assert "semantic_preflight.py" in str(results[0].get("file_path", ""))
def test_classic_search_code_fts_demotes_init_for_filename_match(temp_db_path):
store = SQLiteStore(str(temp_db_path))
repo_id = store.create_repository("/repo", "repo")
store.store_file(
repository_id=repo_id,
path="/repo/mcp_server/setup/__init__.py",
relative_path="mcp_server/setup/__init__.py",
language="python",
)
store.store_file(
repository_id=repo_id,
path="/repo/mcp_server/setup/semantic_preflight.py",
relative_path="mcp_server/setup/semantic_preflight.py",
language="python",
)

Xet Storage Details

Size:
35.7 kB
·
Xet hash:
eebb150334a26bc5ba708333d30ee92019b91f29249d0494864338fa1a27dd50

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.