Adding Specific Files Doesn't Seem To Work?

[DennisTheMenace780]

I anticipate that this is predominantly a user error, but I can't seem to actually add a single file to the chat buffer. If I do something like:

#file: /path/to/my/file.thing

I get back:

[CopilotChat] [Error 14:22:38] /Users/dennis/.local/share/nvim/lazy/CopilotChat.nvim/lua/CopilotChat/init.lua:373: Failed to resolve context: file attempt to a call a nil value.

Which seems strange because that file exists, so i'm not sure how to be selective about which files I want to add into the context. Any ideas? I think this would be really helpful in adding custom file contexts when I start a chat and i'd like to really figure that out.

[deathbeam]

You need to do it without space after :

[DennisTheMenace780]

Ah, yeah I see why this is actually. In the configuration file I was trying to load in a default set of files. So I had the contexts = { file = {'some/default/files} } and that must've been why. But I realize that's not a list of files you want as default contexts.

I had a few other questions around the file search as well:

• Is it possible to truncate the file path? For example, if I have /Users/dennis/projects/main_project/more/files, then all I reaally care about if I am inside of the main_project directory is /more/files, not necessarily the whole path. Maybe that's a telescope configuration thing, or something to be done if customizing how #file: works.
• Is it also possible to exclude files? Again, i'm thinking that might require just customizing how #file: is implemented.

[deathbeam]

1. not atm but i want to look at that in future as we already have notion of main directory in the plugin anyway, but i remember there were some issues with LLMs getting a bit confused
2. Its not possible to exclude via the input (but we are already excluding most binary files), but you can use glob pattern as input

[DennisTheMenace780]

Gotcha, just some additional context for 2, since I think 1 is obvious why that would be helpful.

If I'm working in a large enterprise monorepo project, there can very often be certain areas that I know with confidence I will never need to look at. For example, right now I have 31000 files that Telescope needs to handle, and I could confidently cut that down by 60-70%. That would just make the UX a little nicer, maybe improve performance if we can confidently ignore a lot of files.

[deathbeam]

Well .gitignore files are supported and I believe .ignore files too because its using rg as long as you have it installed if i remember correctly