Questions about Key Mapping, Model Name Display, and Autocompletion for CodeCompanion Command

[bassamsdata]

Hello,

While maximizing my usage of the plugin, I've run into a few questions.

1. When I map the CodeCompanion command to a keymap, it doesn't seem to work. However, it works fine when manually typed in the command line. Here's the mapping I'm using:

   vim.keymap.set({ "n", "v" }, "<leader>ai", "<cmd>CodeCompanion<cr>", { desc = "[A]i [I]nline" })

2. I'd like to display the model name (not the provider) that outputs the data beneath the CodeCompanion text. However, I'm unsure where to retrieve this information. The CodeCompanionChatAdapter event only includes the default model, and I couldn't find anything related to dynamic model names.

   The reason I'm asking is that I want to avoid displaying the settings at the top of the chat buffer when it grows too long. This would reduce key strokes, and since I'm experimenting with multiple models at the moment, it would help streamline things. (Apologies for nitpicking on this.)

3. Is it possible to add autocompletions for the CodeCompanionChat command? It has quite a few subcommands now, and I'd find this very useful. I wanted to submit a PR for it, but I've never worked with completion before.

Thank you!

[S1M0N38]

hi,

1. Did you try to set up your keymap in minimal.lua config in the issue template (you can get that when opening an issue)? maybe you issue is caused by keymaps overlapping.
2. idk. sorry
3. Check out https://github.com/nvim-neorocks/nvim-best-practices?tab=readme-ov-file#speaking_head-user-commands . Here they explain how completion for subcommands can be implemented. For a simplify version you can check my nvim plugin template

[olimorris]

Regarding 1), that's because the command expects some input before it can run. I recommend in the workflow tips to apply an abbreviation such as cc as it's 3 keystrokes before you're typing a prompt. Alternatively, you could use the prompt library shortcut for the custom prompt.

For item 2), this is a little confusing as default_model is poorly named. It can represent the default and the current model. The payload sent from the CodeCompanionChatAdapter event will enable you to access this. Another user had a similar request recently which is how it came to be included in the plugin.

And great shout on 3). Thanks to @S1M0N38 as I hadn't realised you could do this with sub commands.

[bassamsdata]

Thank you all for the responses! the plugin template and the guide for best practices look amazing and easy to follow. I'll give it a try and see if I can get the autocompletion for the subcommands working.

@olimorris, are you talking about schema.model.default? If so, that always points to the default model in the plugin config and doesn’t update to reflect the current model in use (though the adapter name does change).

[olimorris]

Ooooh you're right. I'll add a new command that fires the new model!

[olimorris]

@bassamsdata this should now be added in 5.3.0

[olimorris]

Regarding subcommands, if we were to add them, we'd need to add the slash commands from the prompt_library for full coverage. I'm not convinced their worthwhile adding in. We'll also need to make sure we don't break the legendary.nvim config.

[bassamsdata]

@olimorris Thanks for the quick implementation—it works perfectly!

if we were to add them, we'd need to add the slash commands. We'll also need to make sure we don't break the legendary.nvim config.

Sure, makes sense.

I'm not convinced they're worthwhile adding in.

I'll give it a go and see what I can come up with. If it works, awesome! If not, at least I’ll have learned something new. Plus, I'm working on a module for yanking, so I’m thinking of adding completions for that as well.

[olimorris]

Sounds good, feel free to open up a PR to discuss it if it's going to change the UI or alter the API

[olimorris]

I've added in sub commands as of 8.1.0

[bassamsdata]

that's great, thank you

[bassamsdata]

Just a small nitpick: the completion for both adapters and custom prompts doesn’t seem to recognize user-defined adapters or prompts. After checking the code, it looks like it should be picking them up, since the setup function updates the config—but I’m not sure why it’s not working as expected.

[olimorris]

I tested it with mine and it seems to work. I did push that fix late into the day yesterday so are you on the latest version of the plugin?

[bassamsdata]

Looks like I commented just after the fix went live. It’s working perfectly now. thanks