microsoft/semantic-kernel · error · ServiceResourceNotFoundError

Collection {collection_name} does not exist

Error message

Collection {collection_name} does not exist

What it means

Raised by USearchMemoryStore.get_batch (ServiceResourceNotFoundError) when the target collection is not in `self._collections`, before attempting to resolve keys to labels. Unlike get() (which raises per-key), this guards the whole collection; the message is the plain 'Collection X does not exist'.

Source

Thrown at python/semantic_kernel/connectors/memory_stores/usearch/usearch_memory_store.py:397

            keys=[key],
            with_embeddings=with_embedding,
            dtype=dtype,
        )
        if not result:
            raise ServiceResourceNotFoundError(f"Key '{key}' not found in collection '{collection_name}'")
        return result[0]

    async def get_batch(
        self,
        collection_name: str,
        keys: list[str],
        with_embeddings: bool,
        dtype: ScalarKind = ScalarKind.F32,
    ) -> list[MemoryRecord]:
        """Retrieve a batch of MemoryRecords using their keys."""
        collection_name = collection_name.lower()
        if collection_name not in self._collections:
            raise ServiceResourceNotFoundError(f"Collection {collection_name} does not exist")

        ucollection = self._collections[collection_name]
        labels = [ucollection.embeddings_id_to_label[key] for key in keys if key in ucollection.embeddings_id_to_label]
        if not labels:
            return []
        vectors = ucollection.embeddings_index.get(labels, dtype) if with_embeddings else None

        return pyarrow_table_to_memoryrecords(ucollection.embeddings_data_table.take(pa.array(labels)), vectors)

    async def remove(self, collection_name: str, key: str) -> None:
        """Remove a single MemoryRecord using its key."""
        collection_name = collection_name.lower()
        await self.remove_batch(collection_name=collection_name, keys=[key])
        return

    async def remove_batch(self, collection_name: str, keys: list[str]) -> None:
        """Remove a batch of MemoryRecords using their keys."""
        collection_name = collection_name.lower()

View on GitHub (pinned to c028a0c7dc)

Solutions

  1. Create/load the collection first with create_collection (or set persist_directory at construction).
  2. Guard with does_collection_exist / `'docs' in store._collections` before reading.
  3. Match the collection name exactly (case-insensitive).

Example fix

// before
recs = await store.get_batch('docs', keys, with_embeddings=True)
// after
if not await store.does_collection_exist('docs'):
    recs = []
else:
    recs = await store.get_batch('docs', keys, with_embeddings=True)
Defensive patterns

Strategy: validation

Validate before calling

name = collection_name.lower()
if not await store.does_collection_exist(name):
    return []
return await store.get_batch(name, keys, with_embeddings=True)

Try / catch

from semantic_kernel.exceptions import ServiceResourceNotFoundError
try:
    return await store.get_batch(collection_name, keys, with_embeddings=True)
except ServiceResourceNotFoundError:
    return []

Prevention

When it happens

Trigger: Calling `await store.get_batch('docs', ['k1','k2'], with_embeddings=True)` against a collection that was never created/loaded in this process.

Common situations: Collection name typo or case mismatch (lowercased internally); in-memory store after restart; collection not loaded because persist_directory is unset; reading before any create_collection.

Related errors

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/259d9521457be68e. Report an issue: GitHub.