HuggingFace Embeddings Inference Component

[Nisim1]

Hi community!
I am trying to use HuggingFace Embeddings Inference.

My goal is to use open-source embedding models to perform text embedding.

I am getting an error that says:
"Embedding model client not properly initialized."

When I try to run the same model via Postman, it works!

How can this be resolved?
(The langFlow it's on Docker if it's matter)

Thanks in Advance!

[dosubot[bot]]

Hi @Nisim1! I'm here to help you with any bugs or questions you have. Let's squash those bugs together!

To resolve the "Embedding model client not properly initialized" error when using HuggingFace Embeddings Inference in Docker, you can follow these steps:

1. Validate the Inference Endpoint: Ensure that the inference endpoint URL is correctly formatted and includes both a scheme (e.g., 'http://' or 'https://') and a domain name. This is crucial for the client to connect properly.

2. Check Endpoint Health: Perform a health check on the endpoint by sending a GET request to "{inference_endpoint}/health". Ensure that this endpoint is accessible and returns a status code of 200 (OK).

3. API Key Configuration: Make sure that the API key is correctly configured. For non-local inference endpoints, the API key is required. If the API key is missing, the system will raise a ValueError.

4. Docker Network Configuration: Since the model works via Postman but not in Docker, there might be network configuration issues in Docker. Ensure that the Docker container has network access to the inference endpoint.

5. Error Handling and Logs: Check the logs for any additional error messages that might provide more context. The create_huggingface_embeddings method is wrapped with a retry mechanism, so reviewing the logs can help identify persistent issues.

These steps should help in diagnosing and resolving the issue with the HuggingFace Embeddings Inference component in Docker ^[1].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[Nisim1]

It's not working nothing from what you mention

[anovazzi1]

hey, what's you langflow version and the browser and os you are using?

[KaliZzo]

I tried couple of versions included 1.1.1 and 1.1.0 and I tried from chrome and firefox and also with win10 and win 11
same result on all of them

[jeevankc1718]

same issue @Nisim1 did you solve it?

[KaliZzo]

No

[ArvinRastegar]

✅ Problem:

You're trying to use HuggingFace Embeddings Inference inside LangFlow running in Docker, and you're seeing this error:

"Embedding model client not properly initialized."

Even though the model works fine when tested via Postman, LangFlow throws this error. So what gives?

🛠️ Solution:

The issue is with how the HuggingFaceInferenceAPIEmbeddings class is imported and initialized in the LangFlow HuggingFace Embeddings Inference component.

You’ll need to manually edit the component code as follows:

1. Change the import statement

Replace:

from langchain_community.embeddings.huggingface import HuggingFaceInferenceAPIEmbeddings

With:

from langchain_community.embeddings import HuggingFaceInferenceAPIEmbeddings

2. Update the create_huggingface_embeddings function

Replace:

def create_huggingface_embeddings(
    self, api_key: SecretStr, api_url: str, model_name: str
) -> HuggingFaceInferenceAPIEmbeddings:
    return HuggingFaceInferenceAPIEmbeddings(api_key=api_key, api_url=api_url, model_name=model_name)

With:

def create_huggingface_embeddings(
    self, api_key: SecretStr, api_url: str, model_name: str
) -> HuggingFaceInferenceAPIEmbeddings:
    return HuggingFaceInferenceAPIEmbeddings(api_key=api_key, model_name=model_name)

This change ensures the component is properly initialized with the HuggingFace model.

---

🐳 Bonus (If You’re Running in Docker):

If you’re running everything locally (inside Docker), make sure to set the inference_endpoint using the correct Docker-accessible schema.

For example:

http://<your-container-hostname>:<port>/embed

Instead of the default Hugging Face API endpoint.