Need help with embedding model error in dataset chat

[Arcadix]

Hi Community Team,

I am facing an issue while configuring embedding models in our documentation assistant / dataset chat setup, and I wanted to check whether anyone has seen this before.

Environment:

• Application version: v0.22.1
• Chat works through the UI, but dataset/knowledge-base chat is failing at the embedding step

Issue:
When I try to use embedding models, the system shows errors like below:

*** 404 models/embedding-001 is not found for API version v1beta, or is not supported for embedContent

• Fail to access embedding model(text-embedding-004) using this api key
• models/text-embedding-004 is not found for API version v1beta, or is not supported for embedContent**

What I observed:

• LLM selection appears to work
• The failure seems to happen specifically when the system tries to use the embedding model for dataset retrieval / knowledge-base chat
• I tested with both embedding-001 and text-embedding-004, and both failed

Could you please help clarify:

1. Which embedding models are actually supported in v0.22.1 for dataset/knowledge-base chat?
2. Is there any specific provider configuration required for embeddings?
3. Does this version support Gemini embeddings for embedContent?
4. If OpenAI embeddings are supported, which exact model name should be used?
5. After changing the embedding model, do we need to re-index the dataset?

It would also help if someone can share a known working combination of:

• Provider
• LLM model
• Embedding model
• Any required configuration steps

Thank you.

[Arcadix]

earlier it working fine.

[Arcadix]

Please Suggest the fix.

[Ganesh-403]

@Arcadix The error message:

404 models/embedding-001 is not found for API version v1beta,
or is not supported for embedContent

suggests that the request is reaching Google's API, but the embedding model name or provider configuration doesn't match the endpoint being used, rather than there being an issue with dataset chat itself.

A few things to check:

1. Which Google provider are you using?

   • Gemini Developer API (AI Studio)
   • Vertex AI
   • or an OpenAI-compatible proxy?

   The available embedding models differ depending on the provider.

2. Verify the embedding model name.

   Google's current Gemini API documentation uses models such as:

   • gemini-embedding-001
   • gemini-embedding-2

   for the embedContent API. :contentReference[oaicite:0]{index=0}

   If RAGFlow is sending:

   models/embedding-001
   

   or

   models/text-embedding-004
   

   against the Gemini Developer API, the API may return the 404 you're seeing because those model IDs are not available for that endpoint. :contentReference[oaicite:1]{index=1}

3. If you change the embedding model, re-index the dataset.

   Embeddings generated by different models are generally not compatible with each other. After switching to a different embedding model, it's recommended to reparse/re-index the dataset so all vectors are generated with the same model. :contentReference[oaicite:2]{index=2}

Could you also share:

• Are you using Google AI Studio (Gemini API) or Vertex AI?
• What is the exact embedding model configured in Model Provider?
• Was this working before and then stopped, or is this a fresh v0.22.1 deployment?

That information would help determine whether this is a configuration issue, a model deprecation, or a provider compatibility issue.

If this solves your problem, feel free to mark it as the accepted answer so others can find it easily.