HKUDS/DeepTutor · error · GraphRagNotConfiguredError

No active embedding model with a known dimension. Configure

Error message

No active embedding model with a known dimension. Configure one under Settings → Catalog before creating a GraphRAG knowledge base.

What it means

GraphRagNotConfiguredError from build_settings: an embedding model is active but its dimension (embedding_cfg.dim) is 0/unknown. GraphRAG must know the vector dimension up front to size its vector stores.

Source

Thrown at deeptutor/services/rag/pipelines/graphrag/config.py:269

        from deeptutor.services.embedding import get_embedding_config

        embedding_cfg = get_embedding_config()

    chat_model = getattr(llm_cfg, "model", None)
    embed_model = getattr(embedding_cfg, "model", None)
    embed_dim = int(getattr(embedding_cfg, "dim", 0) or 0)
    if not chat_model:
        raise GraphRagNotConfiguredError(
            "No active chat model. Configure one under Settings → Catalog before "
            "creating a GraphRAG knowledge base."
        )
    if not embed_model:
        raise GraphRagNotConfiguredError(
            "No active embedding model. Configure one under Settings → Catalog "
            "before creating a GraphRAG knowledge base."
        )
    if not embed_dim:
        raise GraphRagNotConfiguredError(
            "No active embedding model with a known dimension. Configure one under "
            "Settings → Catalog before creating a GraphRAG knowledge base."
        )

    embedding_binding = str(getattr(embedding_cfg, "binding", "") or "")
    llm_base = getattr(llm_cfg, "effective_url", None) or getattr(llm_cfg, "base_url", None)
    embed_endpoint = getattr(embedding_cfg, "effective_url", None) or getattr(
        embedding_cfg, "base_url", None
    )
    ensure_graphrag_embedding_transport(embedding_binding, embed_endpoint)
    embed_base = graphrag_embedding_api_base(embedding_binding, embed_endpoint)

    return {
        "completion_models": {
            COMPLETION_MODEL_ID: _completion_model_entry(llm_cfg, api_base=llm_base),
        },
        "embedding_models": {
            EMBEDDING_MODEL_ID: _embedding_model_entry(

View on GitHub (pinned to 3e82f13042)

Solutions

1. Set the embedding dimension explicitly in the embedding profile settings (e.g. 1536 for text-embedding-3-small).
2. Re-select the model from the catalog so its known dimension is populated.
3. Verify with get_embedding_config().dim > 0 before triggering GraphRAG indexing.

Example fix

# before
embedding_cfg.dim = 0
# after
embedding_cfg.dim = 1536

if not int(getattr(get_embedding_config(), "dim", 0) or 0):
    raise ConfigError("embedding profile needs an explicit dimension")

When it happens

Trigger: An embedding profile was created without a dim field (custom/OpenAI-compatible endpoint where the dimension wasn't probed or specified), then GraphRAG settings are built.

Common situations: Manually edited custom embedding profiles missing dimensions; new/unknown embedding models without metadata; profiles migrated from older versions that didn't track dim.

Related errors

• No active embedding model. Configure one under Settings → Ca
• graphrag_embedding_dimension_mismatch
• graphrag_embedding_provider_unsupported
• No active chat model. Configure one under Settings → Catalog
• graphrag_embedding_probe_failed

AI-assisted analysis of HKUDS/DeepTutor@3e82f13042 (2026-08-27). Data as JSON: /api/errors/2d8faa8ee6bab1b9. Report an issue: GitHub.