feature: Allow changing Copilot models on the fly

[repparw]

Feature request

Allow changing copilot models without changing config (maybe as a second parameter on AvanteSwitchProvider copilot ?)

also could be interesting to check for available models somehow (CopilotChat has some kind of fetch_models for :CopilotChatModels, which brings a select)

[Properrr]

Hello! I've implemented a workaround for this case. You can now switch the Copilot model at runtime using <leader>am.

Here's my current avante.nvim configuration:

return {
  {
    "yetone/avante.nvim",
    event = "VeryLazy",
    lazy = false,
    version = false, -- Use "*" for latest release, or false for latest code.
    opts = {
      provider = "copilot",
      copilot = {
        model = "claude-3.7-sonnet",  -- Default model
      },
    },
    build = "make",
    dependencies = {
      "nvim-treesitter/nvim-treesitter",
      "stevearc/dressing.nvim",
      "nvim-lua/plenary.nvim",
      "MunifTanjim/nui.nvim",
      "nvim-tree/nvim-web-devicons",
      "zbirenbaum/copilot.lua",
      {
        "HakonHarnes/img-clip.nvim",
        event = "VeryLazy",
        opts = {
          default = {
            embed_image_as_base64 = false,
            prompt_for_file_name = false,
            drag_and_drop = { insert_mode = true },
            use_absolute_path = true,
          },
        },
      },
      {
        "MeanderingProgrammer/render-markdown.nvim",
        opts = {
          file_types = { "markdown", "Avante" },
        },
        ft = { "markdown", "Avante" },
      },
    },
    config = function(_, opts)
      local available_models = {
        ["claude-3.7-sonnet"] = { model = "claude-3.7-sonnet" },
        ["claude-3.7-sonnet🙅‍♂️🛠️"] = { model = "claude-3.7-sonnet", disable_tools = true },
        ["claude-3.5-sonnet"] = { model = "claude-3.5-sonnet" },
        ["o3-mini-high"] = { model = "o3-mini", reasoning_effort = "high" },
        ["o4-mini-high"] = { model = "o4-mini", reasoning_effort = "high" },
        ["o4-mini-high🙅‍♂️🛠️"] = { model = "o4-mini", reasoning_effort = "high", disable_tools = true },
        ["o3-mini"] = { model = "o3-mini" },
        ["o4-mini"] = { model = "o4-mini" },
        ["4o"] = { model = "gpt-4o" },
        ["4.1"] = { model = "gpt-4.1" },
        ["4.1🙅‍♂️🛠️"] = { model = "gpt-4.1", disable_tools = true },
        ["4.1-mini"] = { model = "gpt-4.1-mini" },
        ["gemini-2.5-pro"] = { model = "gemini-2.5-pro" },
        ["gemini-2.5-pro🙅‍♂️🛠️"] = { model = "gemini-2.5-pro", disable_tools = true },
        ["gemini-2.0-flash"] = { model = "gemini-2.0-flash" },
        ["o2"] = { model = "o2" },
        ["o3"] = { model = "o3" },
        ["o1"] = { model = "o1", reasoning_effort = "high" },
      }

      local function switch_model()
        local model_keys = vim.tbl_keys(available_models)
        vim.ui.select(model_keys, { prompt = "Select Avante Model:" }, function(selected)
          if selected then
            opts.copilot = available_models[selected]
            require("avante").setup(opts)
            print("Switched Copilot model to: " .. selected)
          else
            print("Model selection canceled.")
          end
        end)
      end

      vim.keymap.set("n", "<leader>am", switch_model, { desc = "Avante: Switch Copilot Model" })

      require("avante").setup(opts)
      print("Avante.nvim configured. Use <leader>am to switch models at runtime.")
    end,
  },
}

[repparw]

@Properrr sorry it seems the markdown syntax broke there. still, reading a bit it seems like you're hardcoding all the models currently available there?

Still would be nice to have this as a feature, but I'll take a look at your workaround

[anuramat]

you can pull the models like this

PS: maybe this is worth generalizing, there's a similar request for ollama: #1581; separating models and provders makes sense

[80avin]

fixed in https://github.com/yetone/avante.nvim/pull/2106