[Feature Request]: Add support for local LLM in CLI

[GregorBiswanger]

I would like to request a feature enhancement for the vectra CLI functionality. Specifically, I would like to have the option to use local Large Language Models (LLMs) instead of relying solely on OpenAI's API.

Feature Details:

1. Local LLM Integration:

   • Add support for integrating local LLMs, such as those provided by LM Studio, which allows responses similar to OpenAI.
   • The integration should be flexible to accommodate other LLMs as well.

2. Global Configuration:

   • Implement a global configuration option that allows users to specify their preferred LLM provider.
   • The configuration should include necessary parameters for the chosen LLM, such as API keys or local model paths.

3. CLI Commands:

   • Ensure that CLI commands can easily switch between different LLM providers based on the global configuration.
   • Provide clear documentation and examples on how to set up and use local LLMs within the vectra CLI.

Benefits:

• This feature will provide users with more flexibility and control over their LLM integrations.
• It will allow users to utilize local models, which can be beneficial for privacy, cost, and offline usage scenarios.

Thank you for considering this feature request. I believe it will greatly enhance the functionality and usability of the vectra CLI.

Cheers,
Gregor

[Stevenic]

@GregorBiswanger I recently added support for this... You can use embeddings hosted by any OpenAI compliant server, like llama.cpp. When you create the OpenAIEmbedding class just pass in the ossModel you're using and the ossEndpoint of your server.

[GregorBiswanger]

The idea was to be able to set support for the local LLM via CLI. A profile is defined with a URI, for example for LM Studio localhost:1234 based on OpenAI API structure. You can then index any website and search it semantically via the CLI. An nice to have with vectra... :)