The ability to use OpenAI embeddings with Groq

[S4mpl3r]

Hey there! It would be great if we could use OpenAI embeddings (or any other supported API-based embedding models) with Groq (or any other supported llm). With the current way the code is organized, you can only use OpenAI embeddings with OpenAI models. If I want to use Groq as my main llm, I would have to use Ollama, which is ok if you want to run models locally. But I don't want to install models on my local machine, I would prefer to use OpenAI as my embedder service.

One way to add this is to change the way self.embedder_model is initialized in the AbstractGraph class. Currently, both self.llm_model and self.embedder_model are initialized using one method, self._create_llm(), which kinda makes our options limited. One possible solution is to add another function e.g. self._create_embedder() and completely separate the logic for initialization of llms and embedder models.

[VinciGit00]

Thanks! We will do that

[VinciGit00]

hi, we did it in this file example! link

[S4mpl3r]

Well, that wasn't the inteded solution but nice workaround haha

[VinciGit00]

ok, make a pull request and we will do it

[S4mpl3r]

Alright I'll try to implement my solution and will submit a pull request as soon as I get the chance.

[VinciGit00]

Remember to read the contributions rules :)