Ollama Embeddings is faulty in v1.1 : Error building Component Ollama Embeddings: Could not connect to Ollama API.

[rivalscope]

Bug Description

• It does not connect to API at all ( not working even hardcoded )

• Workaround : Use LMStudio Embeddings component and route to /v1 ( work just fine with ollama even load available models)

For refference please see the image attached.

Reproduction

Try

• Fill as usual the model name and base url.
• Hardcode the model name and base url.

Expected behavior

1 Connect to ollama server
2 Execute the embedding function
3 Respond with the enbedding vectors

Who can help?

No response

Operating System

ubuntu 22.04?! (langflow docker image)

Langflow Version

1.1

Python Version

None

Screenshot

No response

Flow File

No response

[dosubot]

Hello @rivalscope! I'm here to help you with any bugs, questions, or contributions. Let's work together to resolve the issue.

There is a known issue related to the "Could not connect to Ollama API" error when using the Ollama Embeddings component in Langflow v1.1. The suggested solution is to ensure that the base_url is correctly set and that the Ollama API is running and accessible at that URL. The default base_url is http://localhost:11434, so you should verify that the Ollama API is running on your local machine at port 11434. If it's running on a different host or port, update the base_url in the OllamaEmbeddingsComponent configuration accordingly ^[1].

Additionally, a workaround mentioned is to use the LMStudio Embeddings component and route to /v1, which seems to work fine with Ollama and loads available models ^[2].

To continue talking to Dosu, mention @dosu.

[greenmotion]

I am experiencing this as well and someone else posted a similar issue with a potential fix, by removing the temperature setting from the "code" function on the component.