Will local AI models work be implemented?

[Necrodeather]

Will the function of working with local AI models be implemented (if I want to use ollama or lm studio) on my PC or Server?

[jackwotherspoon]

Thanks for the question @Necrodeather

Questions are more for the discussion tab, converting to a discussion 😄

[acoliver]

Google closed all discussions and pull requests recently for multi-model support. I think this is the right decision. Gemini-cli will be focused on being as good as it can be for Gemini. The Gemini Code Assist team working on this tool will put their resources into the ide support and every aspect of making the CLI better.

Downstream you have options:

1. LLxprt Code -https://github.com/acoliver/llxprt-code - a community-based fork that supports LOCAL models and other providers like openrouter / fireworks / openai / anthropic. It also adds advanced features like Claude code-style, to-do lists. It keeps up with the Gemini-CLI codebase with an automerge script that runs and cherry-picks Google's changes. (so it is compatible with 0.1.15)
2. Qwen Code - https://github.com/QwenLM/qwen-code - by the makers of Qwen - supports Qwen models

It's difficult for a company like Google to pursue its own roadmap and aims to incorporate community feedback and direction. Luckily github and LLMs make it possible to do up-to-date community forks!