chroma-core/chroma · error · ValueError

Schema is missing defaults.float_list.vector_index

Error message

Schema is missing defaults.float_list.vector_index

What it means

This ValueError from update_schema_from_collection_configuration indicates an internal invariant broke: while applying a collection-configuration update, the collection's Schema has no defaults.float_list.vector_index (the object describing how the #embedding column is indexed). Every properly created collection with a vector index has this, so hitting it usually means the schema was constructed outside the normal path, is malformed/None, or the collection predates vector-index defaults.

Source

Thrown at chromadb/api/collection_configuration.py:832

) -> "Schema":
    """
    Updates a schema with configuration changes.
    Only updates fields that are present in the configuration update.

    Args:
        schema: The existing Schema object
        configuration: The configuration updates to apply

    Returns:
        Updated Schema object
    """

    # Get the vector index from defaults and #embedding key
    if (
        schema.defaults.float_list is None
        or schema.defaults.float_list.vector_index is None
    ):
        raise ValueError("Schema is missing defaults.float_list.vector_index")

    embedding_key = "#embedding"
    if embedding_key not in schema.keys:
        raise ValueError(f"Schema is missing keys[{embedding_key}]")

    embedding_value_types = schema.keys[embedding_key]
    if (
        embedding_value_types.float_list is None
        or embedding_value_types.float_list.vector_index is None
    ):
        raise ValueError(
            f"Schema is missing keys[{embedding_key}].float_list.vector_index"
        )

    # Update vector index config in both locations
    for vector_index in [
        schema.defaults.float_list.vector_index,
        embedding_value_types.float_list.vector_index,

View on GitHub (pinned to aecdd12c8a)

Solutions

Create a fresh collection and re-ingest the data instead of modifying the malformed one
Upgrade the Chroma server/client together so collections always carry a complete schema before you call modify
If operating a custom build, ensure collection creation assigns defaults.float_list.vector_index; report the issue upstream if a stock server produces this
Check collection metadata/configuration first and only call modify on collections that expose a vector index

Example fix

// before
old = client.get_collection('legacy_from_v0.4')
old.modify(configuration={'hnsw': {'ef_search': 200}})  # schema lacks vector index

// after
new = client.create_collection('legacy_migrated', configuration={'hnsw': {'ef_search': 200}})
new.add(ids=old.get()['ids'], embeddings=old.get(include=['embeddings'])['embeddings'],
        documents=old.get(include=['documents'])['documents'])

Defensive patterns

Strategy: try-catch

Validate before calling

def collection_supports_index_update(collection) -> bool:
    cfg = collection.configuration or {}
    return 'hnsw' in cfg or 'spann' in cfg or cfg.get('embedding_function') is not None

# heuristic client-side check; the authoritative check is the server schema
if collection_supports_index_update(collection):
    collection.modify(configuration=update)

Try / catch

try:
    collection.modify(configuration=update)
except ValueError as e:
    if 'Schema is missing defaults.float_list.vector_index' in str(e):
        # schema malformed: migrate data into a freshly created collection instead
        migrate_to_new_collection(collection)
    else:
        raise

Prevention

Upgrade Chroma client and server together before modifying legacy collections
Create collections with an explicit configuration so schemas are always complete
Avoid modify on collections restored from very old versions; recreate them

When it happens

Trigger: Calling client.modify_collection/collection.modify with an hnsw or spann update where the server-side Schema object lacks defaults.float_list or defaults.float_list.vector_index — e.g. collections restored from very old Chroma versions, hand-built/test schemas, or a schema where defaults were dropped during a migration. The error is raised before any field updates are applied.

Common situations: Persisted databases created by much older Chroma versions being upgraded in place and then modified; test harnesses that fabricate Schema objects directly; bugs in custom server builds or forks that skip default vector-index assignment at collection creation.

Related errors

AI-assisted analysis of chroma-core/chroma@aecdd12c8a (2026-08-16). Data as JSON: /api/errors/541694dd80b62ef5. Report an issue: GitHub.