Using Flowise with infinity as embeddings server

[highmanphil]

Hey, I'm trying to setup https://github.com/FlowiseAI/Flowise with infinity as an embeddings server in the backend. While trying to get everything working, I got into an error namly:
Flowise is using langchain OpenAIEmbeddings and it somehow doesn't send a list of strings, when sending the prompt (it sends the input as a simple string).
With the huggingface/text-inference it was working - possibly they implemented both request types.
Does anyone know any kind of workaround for this kind of issue?

[michaelfeil]

I implemented a langchain version using Infinity embeddings. Its in all langchain packages since around ~0.0.330.
https://python.langchain.com/docs/integrations/text_embedding/infinity

[highmanphil]

Yeah, I already saw that - but the other project is using langchain.ts (is infinity also available in js?)

[michaelfeil]

Sadly no. Also sending list of tokens is just not a thing for non-openai models: #36

[highmanphil]

So, I got it working with a small translation server in the middle
It seams like langchain in javascript just sends either input: str (if you make a query) or list[str] if you embed multiple strings.
My server now only puts the single string into an array with the string as its only element.
Kinda works, but isn't that beautiful