This is a very intriguing and really smart approach. It would be very helpful if the contributors really do not drop this one and keep on upgrading the concept and adding new features like one-click installation, a more robust UI and canvas and other new features like built-in RAG support to keep things interesting and efficient.
can you also add inference time training of the local model so over time it keeps on improving?
Local LLMs, especially small models, have come a long way since then, as have big models like ChatGPT, Gemini and Claude models; the approach to combine the powers of small and big LLMs will reduce the API costs.
This is a very intriguing and really smart approach. It would be very helpful if the contributors really do not drop this one and keep on upgrading the concept and adding new features like one-click installation, a more robust UI and canvas and other new features like built-in RAG support to keep things interesting and efficient.
can you also add inference time training of the local model so over time it keeps on improving?
Local LLMs, especially small models, have come a long way since then, as have big models like ChatGPT, Gemini and Claude models; the approach to combine the powers of small and big LLMs will reduce the API costs.