You can, view the wiki: https://github.com/yetone/avante.nvim/wiki scroll to the bottom of the page

openai = {
  endpoint = "http://127.0.0.1:3000",
  model = "code-gemma",
  temperature = 0,
  max_tokens = 4096,
  ["local"] = true,
},

ollama uses the same calls as the openai api, so this part of the installation instructions should work with ollama, lm-studio, ... any local llm engine that follows the openai api.

Hi, after I made the following changes:

    opts = {
      provider = "openai",
      openai = {
        endpoint = "http://127.0.0.1:11434",
        model = "codegemma",
        temperature = 0,
        max_tokens = 4096,
        ["local"] = true,
      },
    },

I still get prompted to enter the API key when I started NeoVim, and getting errors:

• Failed to setup openai. Avante won't work as expected
• API request failed with status 404

I've tried:

1. Changed the endpoint to http://127.0.0.1:11434/api/chat
2. Use a different model
3. Ensure Ollama is running with

   curl http://localhost:11434/api/chat -d '{                                             
       "model": "codegemma",
       "messages": [
         { "role": "user", "content": "why is the sky blue?" }
       ]
   }'

But none of the above works. Am I supposed to create a custom vendor/provider instead for Ollama?

the endpoint has to be endpoint/v1 (avante will append /chat/completions to it)

provider = "openai",

openai = {
	endpoint = "http://localhost:11434",
	-- endpoint = "http://localhost:11434/v1", -- doesn't work as well
	model = "codegemma:2b",
	temperature = 0,
	max_tokens = 4096,
	["local"] = true,
},
-- ... rest

Still not working for me, it asks for api key

EDIT:

It works, I just put dummy api token and works.. no idea how to disable the prompt for token completely

can you try with #286?

can you try with #286?

now gives an error when I send message in chat, instead of prompting when opening nvim

aight I will add a case for local llm to test this thoroughly, but #295 should address this