add option to preprocess the prompt payload

[albohlabs]

Hi, I could not get my Neovim to talk to my codellama model because the expected format is not the same as the one the plugin produces.

curl -X POST http://localhost:11434/api/generate -d '{
  "model": "codellama",
  "prompt": "Write me a function that outputs the fibonacci sequence"
}'

For this reason I added the preprocess_body option, so that the user of the plugin can easily manipulate the format of the curl post body.

Aaaaaand there is a second commit with some code style changes produced by the lua lsp. Sorry about that. 👼

[Mte90]

For codellama it is requested something with this plugin at the end (apart that PR)?

[albohlabs]

Sorry, I don't think I understand what you mean.

But I saw in your dots that you're also using codellama. What is the curl post body that gen.nvim produces? I'm confused. 🤔

[Mte90]

I was just trying to understand if with codellama once that this PR is approved it is required something else to add in the plugin settings.

[albohlabs]

The preprocessor function has a default implementation that is used and contains the current prompt creation logic. You can find it at https://github.com/albohlabs/gen.nvim/blob/feat/preprocess-body/lua/gen/init.lua#L35

local default_options = {
    ...
    --- Create the curl prompt that will be included in the substutute of $body
    ---
    ---@param prompt string user input and optionally the highlighted code
    ---@return table
    preprocess_body = function(prompt)
        local messages = {}
        table.insert(messages, { role = "user", content = prompt })
        return { messages = messages }
    end,
   ...

This means that when this PR is merged, there is nothing for you to do. Everything should work out of the box.

[David-Kunz]

Hi @albohlabs ,

Sorry for the late reply, is this still needed? I'm a bit confused why codellama has a different shape, I thought Ollama has an OpenAPI compatible API?