infiniflow/ragflow · error · Exception

Cannot search across datasets where some have embedding mode

Error message

Cannot search across datasets where some have embedding models and others do not.

What it means

Raised (as a generic Exception) from get_models in dialog_service when validate_dataset_embedding_models returns an error: the dialog references multiple datasets (kb_ids) where some have an embedding model configured and others have none. Mixed embedding state cannot be searched together, so model resolution aborts before building the embedding LLMBundle.

Source

Thrown at api/db/services/dialog_service.py:359

                yield {"answer": "", "reference": {}, "audio_binary": None, "prompt": "", "created_at": time.time(), "final": False, **flags}
                continue
            yield {"answer": value, "reference": {}, "audio_binary": tts(tts_mdl, value), "prompt": "", "created_at": time.time(), "final": False}
    else:
        if model_config["model_type"] == "chat":
            answer = await chat_mdl.async_chat(system_prompt, msg, dialog.llm_setting)
        else:
            answer = await chat_mdl.async_chat(system_prompt, msg, dialog.llm_setting, images=image_files)
        user_content = msg[-1].get("content", "[content not available]")
        logging.debug("User: {}|Assistant: {}".format(user_content, answer))
        yield {"answer": answer, "reference": {}, "audio_binary": tts(tts_mdl, answer), "prompt": "", "created_at": time.time()}


def get_models(dialog, trace_context=None, langfuse_session_id=None):
    embd_mdl, chat_mdl, rerank_mdl, tts_mdl = None, None, None, None
    kbs = KnowledgebaseService.get_by_ids(dialog.kb_ids)
    err = validate_dataset_embedding_models(kbs)
    if err:
        raise Exception(err)

    if kbs and kbs[0].embd_id:
        embd_owner_tenant_id = kbs[0].tenant_id
        embd_model_config = resolve_model_config(embd_owner_tenant_id, LLMType.EMBEDDING, kbs[0].embd_id)
        embd_mdl = LLMBundle(embd_owner_tenant_id, embd_model_config, trace_context=trace_context, langfuse_session_id=langfuse_session_id)
        if not embd_mdl:
            raise LookupError("Embedding model(%s) not found" % kbs[0].embd_id)

    if dialog.llm_id:
        if dialog.tenant_llm_id:
            try:
                chat_model_config = get_model_config_by_id(dialog.tenant_id, LLMType.CHAT, dialog.tenant_llm_id)
            except LookupError:
                chat_model_config = resolve_model_config(dialog.tenant_id, LLMType.CHAT, dialog.llm_id)
        else:
            chat_model_config = resolve_model_config(dialog.tenant_id, LLMType.CHAT, dialog.llm_id)
    else:
        chat_model_config = get_tenant_default_model_by_type(dialog.tenant_id, LLMType.CHAT)

View on GitHub (pinned to 554fb1133a)

Solutions

Set an embedding model on the un-embedded dataset(s) — it must match the others (see error 448).
Remove the un-embedded dataset from the dialog's kb_ids.
Re-create the empty dataset selecting the same embedding model as the dialog's other datasets, then swap it in.

Example fix

# before
dialog.kb_ids = [kb_with_embd.id, kb_without_embd.id]
# after: give every dataset the same embedding model
kb_without_embd.embd_id = kb_with_embd.embd_id
kb_without_embd.save()

Defensive patterns

Strategy: validation

Validate before calling

from api.db.services.knowledgebase_service import KnowledgebaseService, validate_dataset_embedding_models

kbs = KnowledgebaseService.get_by_ids(dialog['kb_ids'])
err = validate_dataset_embedding_models(kbs)
if err:
    # fix datasets before opening the chat session
    raise ValueError(err)

Try / catch

try:
    models = get_models(dialog)
except Exception as e:
    if 'embedding models' in str(e):
        show_dataset_config_error(dialog.kb_ids)  # guide user to dataset settings
    else:
        raise

Prevention

Require an embedding model at dataset creation; disallow empty embd_id.
When editing a dialog's datasets, re-run validate_dataset_embedding_models on every change.

When it happens

Trigger: Chat/ask request (or any flow calling get_models) against a dialog whose kb_ids includes at least one dataset with embd_id set and at least one with empty embd_id.

Common situations: Attaching a newly created dataset that never had an embedding model selected to an existing dialog with embedded datasets; a dataset whose embedding model was cleared; migrating dialogs that accumulated datasets over time.

Related errors

AI-assisted analysis of infiniflow/ragflow@554fb1133a (2026-08-15). Data as JSON: /api/errors/9c626562d8e0c379. Report an issue: GitHub.