Closed SmokeShine closed 3 months ago
You can use my fork to use the KoboldCpp API. My fork: https://github.com/Topping1/AskKobold/tree/main KoboldCpp: https://github.com/LostRuins/koboldcpp
Is it possible to use llama openai compatibility ? ggerganov/llama.cpp#795
What changes would be required?
Using @Topping1 's fork may be easiest, but forking and pointing to a server running a spec mimicking openai's isn't too challenging. I could imagine a preference option where one could pick their model or supply an openAI style compatible API endpoint of their own for local LLMs.
Using @Topping1 's fork may be easiest, but forking and pointing to a server running a spec mimicking openai's isn't too challenging. I could imagine a preference option where one could pick their model or supply an openAI style compatible API endpoint of their own for local LLMs.
I believe that making both apibase and module configurable options can meet customization needs and make this project more flexible.
@eMUQI Yes this would be great, and not too challenging. Usually all that's needed is to change the base URL to an OpenAI API clone.
If I added this as an option would you rather see this in the config file like the API key or some form of on device configuration?
@eMUQI Yes this would be great, and not too challenging. Usually all that's needed is to change the base URL to an OpenAI API clone.
If I added this as an option would you rather see this in the config file like the API key or some form of on device configuration?
@drewbaumann Thank you for considering adding this feature. Personally, being able to configure it through a configuration file is sufficient for me. I just successfully used a third-party OpenAI proxy service with simple code modifications; everything works well. This is a fantastic plugin—thank you! However, for those who aren't very familiar with coding, having the option to configure both apibase and module alongside the API key would be great.
@eMUQI I just created this PR which should address your needs: #12
Is it possible to use llama openai compatibility ? https://github.com/ggerganov/llama.cpp/discussions/795
What changes would be required?