Request for Help: CrewAI RagTool Hugging Face LLM/Embedder Setup Error (Want to Huggingface key))

Short Course Q&A ACP: Agent Communication Protocol
,
1

Hi DeepLearning AI Community,

I’m working through the CrewAI RAG tool agent setup as part of ACP: Agent Communication Protocal course, but I’ve hit a roadblock and would appreciate some guidance.

Context:

  • I do not have an OpenAI API key, so I am trying to use Hugging Face for both the LLM and the embedding model.
  • My goal is to use the DeepSeek RL Distill 70B model from Hugging Face as the LLM, and a BAAI BGE model for embeddings.

My config looks like this:

python

config = {
    "llm": {
        "provider": "huggingface",
        "config": {
            "model": "deepseek-ai/deepseek-llm-rl-distilled-70b",
            "api_key": HF_key
        }
    },
    "embedding_model": {
        "provider": "huggingface",
        "config": {
            "model": "BAAI/bge-small-en-v1.5",
            "api_key": HF_key
        }
    }
}

The error I am getting:

text

ModuleNotFoundError: The required dependencies for HuggingFaceHub are not installed. Please install with `pip install langchain_huggingface`

My Doubts & Questions:

  • Is my config correct for using Hugging Face models with CrewAI’s RagTool?
  • For the DeepSeek LLM, is "deepseek-ai/deepseek-llm-rl-distilled-70b" the right model string for Hugging Face?
  • How do I know which embedding models are compatible with the DeepSeek LLM for RAG tasks? Is BAAI/bge-small-en-v1.5 a good choice, or is there a recommended pairing?
  • Are there any other dependencies or setup steps I might be missing (besides installing langchain_huggingface)?
  • If anyone has a working example of this setup (using Hugging Face for both LLM and embedder in CrewAI RAG), could you please share it?

Error Stage:

  • The error occurs immediately when initializing the RagTool with the config above—before any data is loaded or agents are run.
    image
    image1522×512 161 KB

Any advice, clarifications, or pointers to documentation/examples would be greatly appreciated!
Thank you in advance.

2

Since you are trying to recreate this, you have to iterate I’m afraid. Try to install it as requested. Also, try reading the documentation on langchain: Hugging Face | 🦜️🔗 LangChain
You have to check for any further error and address as you see. Also, consider using Gemini too. You can actually run inference using the free tier. Read the crewai documentation here too: LLMs - CrewAI

1 Like