rag-ai-backend-embeddings-openai: configurable URL of the openai endpoint

[neurohrer-mms]

Feature Suggestion

to connect to some selfhosted or connecting to a openai compatible endpoint we need to have a possibility to add this url in the config (optional).

Possible Implementation

1. add optional value to configuration
2. pass the configuration to the OpenAI see https://js.langchain.com/v0.2/docs/integrations/text_embedding/openai

[fjudith]

Hello @neurohrer-mms ,

I'm currently looking to develop a similar feature, but the langchain LangchainJS library does not seems to support endpoint customization embbedings.ts, BUT the OpenAI version does embeddings.py.

[neurohrer-mms]

i think it does. the baseURL property of the configuration parameter (type is ClientOptions). https://v02.api.js.langchain.com/interfaces/_langchain_openai.ClientOptions.html

https://js.langchain.com/v0.2/docs/integrations/text_embedding/openai/#custom-urls

[vmahe35]

Hello,
I am also interested by this feature because I am using a self hosted embeddings with OpenAI compliant interface (provided via VLLM).
Thanks

[pc-bob]

This should be complete now @punkle