mem0ai/mem0 · error · ValueError

Collection not initialized. Call create_col first.

Error message

Collection not initialized. Call create_col first.

What it means

Raised by FAISS.insert() when self.index is None, i.e. no collection was created or loaded before writing. Mem0's FAISS store requires create_col() (or a successful load) to build the underlying faiss.Index before any vectors can be added. ValueError.

Source

Thrown at mem0/vector_stores/faiss.py:350

        return self

    def insert(
        self,
        vectors: List[list],
        payloads: Optional[List[Dict]] = None,
        ids: Optional[List[str]] = None,
    ):
        """
        Insert vectors into a collection.

        Args:
            vectors (List[list]): List of vectors to insert.
            payloads (Optional[List[Dict]], optional): List of payloads corresponding to vectors. Defaults to None.
            ids (Optional[List[str]], optional): List of IDs corresponding to vectors. Defaults to None.
        """
        if self.index is None:
            raise ValueError("Collection not initialized. Call create_col first.")

        if ids is None:
            ids = [str(uuid.uuid4()) for _ in range(len(vectors))]

        if payloads is None:
            payloads = [{} for _ in range(len(vectors))]

        if len(vectors) != len(ids) or len(vectors) != len(payloads):
            raise ValueError("Vectors, payloads, and IDs must have the same length")

        vectors_np = np.array(vectors, dtype=np.float32)

        if self._should_normalize():
            faiss.normalize_L2(vectors_np)

        self.index.add(vectors_np)

        starting_idx = len(self.index_to_id)

View on GitHub (pinned to 001c235229)

Solutions

Call create_col(vectors dimension) once before the first insert
Prefer the Memory/MemoryConfig pipeline, which initializes the collection for you
If loading an existing store, check logs for 'Failed to load FAISS index' — the initializer may have silently reset state, meaning the on-disk data needs fixing first

Example fix

# before
vs.insert(vectors=[...])  # ValueError: Collection not initialized

# after
vs.create_col(len(embedding))
vs.insert(vectors=[...], payloads=[...], ids=[...])

Defensive patterns

Strategy: validation

Validate before calling

if vs.index is None:
    vs.create_col(dimension)  # dimension = len(embedding)

Prevention

Always call create_col(dim) before the first insert
Watch for 'Failed to load FAISS index' warnings — the init fallback leaves index None
Prefer the Memory API, which handles collection creation

When it happens

Trigger: Constructing FAISS with no existing on-disk store (fresh path) and calling insert()/add() before create_col(); or a load failure that silently left self.index None (the broad except in __init__ resets state instead of raising).

Common situations: Using the FAISS store directly instead of through Memory, which calls create_col automatically; a corrupted load path that swallowed an exception (logger.warning then empty dicts), leaving the object half-initialized.

Related errors

AI-assisted analysis of mem0ai/mem0@001c235229 (2026-08-15). Data as JSON: /api/errors/9de637bb5fd3e408. Report an issue: GitHub.