fix: block ChromaDB 1.5.7 segfault triggers and search None crash

jphein · claude · jphein · commit d3a2d2206e9a · 2026-04-18T09:08:45.000-07:00
Two in-code triggers for ChromaDB 1.5.7 Rust-bindings segfaults were
collapsed under the fork's generic "call mempalace_reconnect" workaround.
Both are fixable:

- get_or_create_collection(name, metadata={"hnsw:space":"cosine"}) on an
  existing collection with differing metadata crashes the Rust compactor.
  Fetch first; only pass metadata on fresh creation. Applied in
  ChromaBackend.get_collection and mcp_server._get_collection.

- Opening a Python sqlite3 connection against the live chroma.sqlite3
  corrupts the next PersistentClient call. _fix_blob_seq_ids now writes
  a .blob_seq_ids_migrated marker after success and skips the sqlite
  open on subsequent runs; already-migrated palaces can touch the
  marker to opt into the fast path.

Also:
- searcher.print-path guards against None drawer metadata (was raising
  AttributeError mid-render after partial results).
- hook scripts marked executable.
- gitignore the .claude-plugin/venv symlink used to point hook
  auto-detection at the repo's venv.

Co-Authored-By: Claude Opus 4.7 (1M context) &lt;noreply@anthropic.com&gt;
diff --git a/.claude-plugin/hooks/mempal-precompact-hook.sh b/.claude-plugin/hooks/mempal-precompact-hook.sh
diff --git a/.claude-plugin/hooks/mempal-stop-hook.sh b/.claude-plugin/hooks/mempal-stop-hook.sh
diff --git a/.gitignore b/.gitignore
@@ -33,6 +33,7 @@ coverage.xml
 # Virtual environments
 .venv/
 venv/
+.claude-plugin/venv
 
 # ChromaDB local data
 *.sqlite3-journal
diff --git a/mempalace/backends/chroma.py b/mempalace/backends/chroma.py
@@ -11,6 +11,9 @@
 logger = logging.getLogger(__name__)
 
 
+_BLOB_FIX_MARKER = ".blob_seq_ids_migrated"
+
+
 def _fix_blob_seq_ids(palace_path: str):
     """Fix ChromaDB 0.6.x -> 1.5.x migration bug: BLOB seq_ids -> INTEGER.
 
@@ -20,25 +23,40 @@ def _fix_blob_seq_ids(palace_path: str):
     type INTEGER) is not compatible with SQL type BLOB".
 
     Must run BEFORE PersistentClient is created (the compactor fires on init).
+
+    Opening a Python sqlite3 connection against a ChromaDB 1.5.x WAL-mode
+    database leaves state that segfaults the next PersistentClient call.
+    After the migration has run once successfully, a marker file is written
+    so subsequent opens skip the sqlite connection entirely.
     """
     db_path = os.path.join(palace_path, "chroma.sqlite3")
     if not os.path.isfile(db_path):
         return
+    marker = os.path.join(palace_path, _BLOB_FIX_MARKER)
+    if os.path.isfile(marker):
+        return
     try:
         with sqlite3.connect(db_path) as conn:
+            table_rows: dict = {}
             for table in ("embeddings", "max_seq_id"):
                 try:
                     rows = conn.execute(
                         f"SELECT rowid, seq_id FROM {table} WHERE typeof(seq_id) = 'blob'"
                     ).fetchall()
                 except sqlite3.OperationalError:
                     continue
-                if not rows:
-                    continue
+                if rows:
+                    table_rows[table] = rows
+            for table, rows in table_rows.items():
                 updates = [(int.from_bytes(blob, byteorder="big"), rowid) for rowid, blob in rows]
                 conn.executemany(f"UPDATE {table} SET seq_id = ? WHERE rowid = ?", updates)
                 logger.info("Fixed %d BLOB seq_ids in %s", len(updates), table)
             conn.commit()
+        try:
+            with open(marker, "w", encoding="utf-8") as f:
+                f.write("migrated\n")
+        except OSError:
+            pass
     except Exception:
         logger.exception("Could not fix BLOB seq_ids in %s", db_path)
 
@@ -125,9 +143,15 @@ def get_collection(self, palace_path: str, collection_name: str, create: bool =
 
         client = self._client(palace_path)
         if create:
-            collection = client.get_or_create_collection(
-                collection_name, metadata={"hnsw:space": "cosine"}
-            )
+            # ChromaDB 1.5.x segfaults when get_or_create_collection is called
+            # with metadata that differs from an existing collection's metadata.
+            # Fetch first; only pass hnsw:space when actually creating fresh.
+            try:
+                collection = client.get_collection(collection_name)
+            except Exception:
+                collection = client.create_collection(
+                    collection_name, metadata={"hnsw:space": "cosine"}
+                )
         else:
             collection = client.get_collection(collection_name)
         return ChromaCollection(collection)
diff --git a/mempalace/mcp_server.py b/mempalace/mcp_server.py
@@ -202,11 +202,16 @@ def _get_collection(create=False):
     try:
         client = _get_client()
         if create:
-            _collection_cache = ChromaCollection(
-                client.get_or_create_collection(
+            # ChromaDB 1.5.x segfaults when get_or_create_collection is called
+            # with metadata that differs from an existing collection's metadata.
+            # Fetch first; only pass hnsw:space when actually creating fresh.
+            try:
+                raw_col = client.get_collection(_config.collection_name)
+            except Exception:
+                raw_col = client.create_collection(
                     _config.collection_name, metadata={"hnsw:space": "cosine"}
                 )
-            )
+            _collection_cache = ChromaCollection(raw_col)
             _metadata_cache = None
             _metadata_cache_time = 0
         elif _collection_cache is None:
diff --git a/mempalace/searcher.py b/mempalace/searcher.py
@@ -270,6 +270,7 @@ def search(query: str, palace_path: str, wing: str = None, room: str = None, n_r
 
     for i, (doc, meta, dist) in enumerate(zip(docs, metas, dists), 1):
         similarity = round(max(0.0, 1 - dist), 3)
+        meta = meta or {}
         source = Path(meta.get("source_file", "?")).name
         wing_name = meta.get("wing", "?")
         room_name = meta.get("room", "?")
