Support for Ollama as a local LLM backend

[jjhaggar]

Hi! I wanted to suggest adding support for Ollama (https://github.com/ollama/ollama) as an alternative backend. This would allow users to run the gaming agent entirely offline, avoiding both the need for an internet connection and the costs associated with API-based models (OpenAI, Anthropic, Google, etc.).

Since Ollama supports vision-language models like LLaVA-1.5, Qwen-VL and others, it could be a great option for local deployment.

Thank you for taking the time to read this! :)

[snyhlxde1]

Great point! We are planning to add local serving backends as well, including Ollama, vLLM etc. Stay tuned!

Meanwhile, we also welcome community contribution : )

[ep150de]

I also recommend and suggest the support for a local only backend so you no longer have to require the API support + cost, rather than using Ollama, there's another backend using llama cpp

I wish I had more time to factor this into your project but here is a guide to reference -> https://blog.steelph0enix.dev/posts/llama-cpp-guide/