fix(core): handle SQLite and Windows semantic regressions

src/basic_memory/repository/search_repository_base.py

@@ -451,21 +451,36 @@ def _compose_row_source_text(self, row) -> str:
         return "\n\n".join(part for part in row_parts if part)
 
     def _build_chunk_records(self, rows) -> list[dict[str, str]]:
-        records: list[dict[str, str]] = []
+        records_by_key: dict[str, dict[str, str]] = {}
+        duplicate_chunk_keys = 0
         for row in rows:
             source_text = self._compose_row_source_text(row)
             chunks = self._split_text_into_chunks(source_text)
             for chunk_index, chunk_text in enumerate(chunks):
                 chunk_key = f"{row.type}:{row.id}:{chunk_index}"
                 source_hash = hashlib.sha256(chunk_text.encode("utf-8")).hexdigest()
-                records.append(
-                    {
-                        "chunk_key": chunk_key,
-                        "chunk_text": chunk_text,
-                        "source_hash": source_hash,
-                    }
-                )
-        return records
+                # Trigger: SQLite FTS5 can accumulate duplicate logical rows for the
+                # same search_index id because it does not enforce relational uniqueness.
+                # Why: duplicate chunk keys would schedule duplicate writes for the same
+                # chunk row and eventually trip UNIQUE(rowid) in search_vector_embeddings.
+                # Outcome: collapse chunk work to one deterministic record per chunk key.
+                if chunk_key in records_by_key:
+                    duplicate_chunk_keys += 1
+                records_by_key[chunk_key] = {
+                    "chunk_key": chunk_key,
+                    "chunk_text": chunk_text,
+                    "source_hash": source_hash,
+                }
+
+        if duplicate_chunk_keys:
+            logger.warning(
+                "Collapsed duplicate vector chunk keys before embedding sync: "
+                "project_id={project_id} duplicate_chunk_keys={duplicate_chunk_keys}",
+                project_id=self.project_id,
+                duplicate_chunk_keys=duplicate_chunk_keys,
+            )
+
+        return list(records_by_key.values())
 
     # --- Text splitting ---
 

src/basic_memory/services/project_service.py

@@ -11,6 +11,7 @@
 
 from loguru import logger
 from sqlalchemy import text
+from sqlalchemy.exc import OperationalError as SAOperationalError
 
 from basic_memory.models import Project
 from basic_memory.repository.project_repository import ProjectRepository
@@ -1004,56 +1005,81 @@ async def get_embedding_status(self, project_id: int) -> EmbeddingStatus:
         )
         total_indexed_entities = si_result.scalar() or 0
 
-        chunks_result = await self.repository.execute_query(
-            text("SELECT COUNT(*) FROM search_vector_chunks WHERE project_id = :project_id"),
-            {"project_id": project_id},
-        )
-        total_chunks = chunks_result.scalar() or 0
-
-        entities_with_chunks_result = await self.repository.execute_query(
-            text(
-                "SELECT COUNT(DISTINCT entity_id) FROM search_vector_chunks "
-                "WHERE project_id = :project_id"
-            ),
-            {"project_id": project_id},
-        )
-        total_entities_with_chunks = entities_with_chunks_result.scalar() or 0
-
-        # Embeddings count — join pattern differs between SQLite and Postgres
-        if is_postgres:
-            embeddings_sql = text(
-                "SELECT COUNT(*) FROM search_vector_chunks c "
-                "JOIN search_vector_embeddings e ON e.chunk_id = c.id "
-                "WHERE c.project_id = :project_id"
-            )
-        else:
-            embeddings_sql = text(
-                "SELECT COUNT(*) FROM search_vector_chunks c "
-                "JOIN search_vector_embeddings e ON e.rowid = c.id "
-                "WHERE c.project_id = :project_id"
+        try:
+            chunks_result = await self.repository.execute_query(
+                text("SELECT COUNT(*) FROM search_vector_chunks WHERE project_id = :project_id"),
+                {"project_id": project_id},
             )
+            total_chunks = chunks_result.scalar() or 0
 
-        embeddings_result = await self.repository.execute_query(
-            embeddings_sql, {"project_id": project_id}
-        )
-        total_embeddings = embeddings_result.scalar() or 0
+            entities_with_chunks_result = await self.repository.execute_query(
+                text(
+                    "SELECT COUNT(DISTINCT entity_id) FROM search_vector_chunks "
+                    "WHERE project_id = :project_id"
+                ),
+                {"project_id": project_id},
+            )
+            total_entities_with_chunks = entities_with_chunks_result.scalar() or 0
+
+            # Embeddings count — join pattern differs between SQLite and Postgres
+            if is_postgres:
+                embeddings_sql = text(
+                    "SELECT COUNT(*) FROM search_vector_chunks c "
+                    "JOIN search_vector_embeddings e ON e.chunk_id = c.id "
+                    "WHERE c.project_id = :project_id"
+                )
+            else:
+                embeddings_sql = text(
+                    "SELECT COUNT(*) FROM search_vector_chunks c "
+                    "JOIN search_vector_embeddings e ON e.rowid = c.id "
+                    "WHERE c.project_id = :project_id"
+                )
 
-        # Orphaned chunks (chunks without embeddings — indicates interrupted indexing)
-        if is_postgres:
-            orphan_sql = text(
-                "SELECT COUNT(*) FROM search_vector_chunks c "
-                "LEFT JOIN search_vector_embeddings e ON e.chunk_id = c.id "
-                "WHERE c.project_id = :project_id AND e.chunk_id IS NULL"
+            embeddings_result = await self.repository.execute_query(
+                embeddings_sql, {"project_id": project_id}
             )
-        else:
-            orphan_sql = text(
-                "SELECT COUNT(*) FROM search_vector_chunks c "
-                "LEFT JOIN search_vector_embeddings e ON e.rowid = c.id "
-                "WHERE c.project_id = :project_id AND e.rowid IS NULL"
+            total_embeddings = embeddings_result.scalar() or 0
+
+            # Orphaned chunks (chunks without embeddings — indicates interrupted indexing)
+            if is_postgres:
+                orphan_sql = text(
+                    "SELECT COUNT(*) FROM search_vector_chunks c "
+                    "LEFT JOIN search_vector_embeddings e ON e.chunk_id = c.id "
+                    "WHERE c.project_id = :project_id AND e.chunk_id IS NULL"
+                )
+            else:
+                orphan_sql = text(
+                    "SELECT COUNT(*) FROM search_vector_chunks c "
+                    "LEFT JOIN search_vector_embeddings e ON e.rowid = c.id "
+                    "WHERE c.project_id = :project_id AND e.rowid IS NULL"
+                )
+
+            orphan_result = await self.repository.execute_query(
+                orphan_sql, {"project_id": project_id}
             )
+            orphaned_chunks = orphan_result.scalar() or 0
+        except SAOperationalError as exc:
+            # Trigger: sqlite_master can list vec0 virtual tables even when sqlite-vec
+            # is not loaded in the current Python runtime.
+            # Why: project info should degrade gracefully instead of crashing on stats queries.
+            # Outcome: report vector tables as unavailable and point the user to install the
+            # missing dependency before rebuilding embeddings.
+            if is_postgres or "no such module: vec0" not in str(exc).lower():
+                raise
 
-        orphan_result = await self.repository.execute_query(orphan_sql, {"project_id": project_id})
-        orphaned_chunks = orphan_result.scalar() or 0
+            return EmbeddingStatus(
+                semantic_search_enabled=True,
+                embedding_provider=provider,
+                embedding_model=model,
+                embedding_dimensions=dimensions,
+                total_indexed_entities=total_indexed_entities,
+                vector_tables_exist=False,
+                reindex_recommended=True,
+                reindex_reason=(
+                    "SQLite vector tables exist but sqlite-vec is unavailable in this Python "
+                    "environment — install/update basic-memory, then run: bm reindex --embeddings"
+                ),
+            )
 
         # --- Reindex recommendation logic (priority order) ---
         reindex_recommended = False

src/basic_memory/utils.py

@@ -66,6 +66,7 @@ def __str__(self) -> str: ...
 # In type annotations, use Union[Path, str] instead of FilePath for now
 # This preserves compatibility with existing code while we migrate
 FilePath = Union[Path, str]
+WINDOWS_LOG_FILE_RETENTION = 5
 
 
 def generate_permalink(file_path: Union[Path, str, PathLike], split_extension: bool = True) -> str:
@@ -250,7 +251,7 @@ def setup_logging(
     log_to_file: bool = False,
     log_to_stdout: bool = False,
     structured_context: bool = False,
-) -> None:  # pragma: no cover
+) -> None:
     """Configure logging with explicit settings.
 
     This function provides a simple, explicit interface for configuring logging.
@@ -273,8 +274,14 @@ def setup_logging(
 
     # Add file handler with rotation
     if log_to_file:
-        log_path = Path.home() / ".basic-memory" / "basic-memory.log"
+        # Trigger: Windows does not allow renaming an open file held by another process.
+        # Why: multiple basic-memory processes can share the same log directory at once.
+        # Outcome: use per-process log files on Windows so log rotation stays local.
+        log_filename = f"basic-memory-{os.getpid()}.log" if os.name == "nt" else "basic-memory.log"
+        log_path = Path.home() / ".basic-memory" / log_filename
         log_path.parent.mkdir(parents=True, exist_ok=True)
+        if os.name == "nt":
+            _cleanup_windows_log_files(log_path.parent, log_path.name)
         # Keep logging synchronous (enqueue=False) to avoid background logging threads.
         # Background threads are a common source of "hang on exit" issues in CLI/test runs.
         logger.add(
@@ -308,6 +315,28 @@ def setup_logging(
     logging.getLogger("watchfiles.main").setLevel(logging.WARNING)
 
 
+def _cleanup_windows_log_files(log_dir: Path, current_log_name: str) -> None:
+    """Trim stale per-process Windows log files so the directory stays bounded."""
+    stale_logs = [
+        path
+        for path in log_dir.glob("basic-memory-*.log*")
+        if path.is_file() and path.name != current_log_name
+    ]
+
+    if len(stale_logs) <= WINDOWS_LOG_FILE_RETENTION - 1:
+        return
+
+    # Trigger: per-process log filenames avoid Windows rename contention but fragment retention.
+    # Why: loguru retention applies per sink, not across the whole basic-memory log directory.
+    # Outcome: keep only the newest stale PID logs so repeated CLI/server launches stay bounded.
+    stale_logs.sort(key=lambda path: path.stat().st_mtime, reverse=True)
+    for stale_log in stale_logs[WINDOWS_LOG_FILE_RETENTION - 1 :]:
+        try:
+            stale_log.unlink()
+        except OSError:
+            logger.debug("Failed to delete stale Windows log file: {path}", path=stale_log)
+
+
 def parse_tags(tags: Union[List[str], str, None]) -> List[str]:
     """Parse tags from various input formats into a consistent list.
 

tests/repository/test_semantic_search_base.py

@@ -237,6 +237,29 @@ def test_chunk_key_includes_row_id(self):
         records = self.repo._build_chunk_records(rows)
         assert any("99" in r["chunk_key"] for r in records)
 
+    def test_duplicate_rows_collapse_to_unique_chunk_keys(self):
+        rows = [
+            _make_row(
+                row_type=SearchItemType.ENTITY.value,
+                title="Spec",
+                permalink="spec",
+                content_snippet="shared content",
+                row_id=77,
+            ),
+            _make_row(
+                row_type=SearchItemType.ENTITY.value,
+                title="Spec",
+                permalink="spec",
+                content_snippet="shared content",
+                row_id=77,
+            ),
+        ]
+
+        records = self.repo._build_chunk_records(rows)
+
+        assert len(records) == 1
+        assert records[0]["chunk_key"] == "entity:77:0"
+
 
 # --- SQLite SemanticSearchDisabledError ---
 

tests/services/test_project_service_embedding_status.py

@@ -5,6 +5,7 @@
 
 import pytest
 from sqlalchemy import text
+from sqlalchemy.exc import OperationalError as SAOperationalError
 
 from basic_memory.schemas.project_info import EmbeddingStatus
 from basic_memory.services.project_service import ProjectService
@@ -142,6 +143,46 @@ async def test_embedding_status_orphaned_chunks(
     assert "orphaned chunks" in (status.reindex_reason or "")
 
 
+@pytest.mark.asyncio
+async def test_embedding_status_handles_sqlite_vec_unavailable(
+    project_service: ProjectService, test_graph, test_project
+):
+    """Unreadable vec0 tables should degrade to unavailable status instead of crashing."""
+    # Trigger: Postgres test matrix executes the same unit suite.
+    # Why: sqlite-vec loading failures are specific to SQLite virtual tables, not Postgres joins.
+    # Outcome: keep the regression focused on the backend that can actually hit this path.
+    if _is_postgres():
+        pytest.skip("sqlite-vec unavailable handling is SQLite-specific.")
+
+    original_execute_query = project_service.repository.execute_query
+
+    async def _execute_query_with_vec0_failure(query, params):
+        query_text = str(query)
+        if "JOIN search_vector_embeddings" in query_text:
+            raise SAOperationalError(query_text, params, Exception("no such module: vec0"))
+        return await original_execute_query(query, params)
+
+    with patch.object(
+        type(project_service),
+        "config_manager",
+        new_callable=lambda: property(
+            lambda self: _config_manager_with(semantic_search_enabled=True)
+        ),
+    ):
+        with patch.object(
+            project_service.repository,
+            "execute_query",
+            side_effect=_execute_query_with_vec0_failure,
+        ):
+            status = await project_service.get_embedding_status(test_project.id)
+
+    assert status.semantic_search_enabled is True
+    assert status.total_indexed_entities > 0
+    assert status.vector_tables_exist is False
+    assert status.reindex_recommended is True
+    assert "sqlite-vec is unavailable" in (status.reindex_reason or "")
+
+
 @pytest.mark.asyncio
 async def test_embedding_status_healthy(project_service: ProjectService, test_graph, test_project):
     """When all entities have embeddings, no reindex recommended."""