A Neovim plugin to help you work out what the fudge that diagnostic means and how to fix it!
wtf.nvim provides faster and more efficient ways of working with the buffer line's diagnostic messages by redirecting them to your favourite resources straight from Neovim.
Works with any language that has LSP support in Neovim.
Use the power of ChatGPT to provide you with explanations and solutions for how to fix diagnostics, custom tailored to the code responsible for them.
Wtf.mp4
Why spend time copying and pasting, or worse yet, typing out diagnostic messages, when you can open a search for them in Google, Stack Overflow and more, directly from Neovim?
wtf-search.mp4
In order to use the AI functionality, set the environment variable OPENAI_API_KEY to your openai api key (the search functionality will still work without it).
Install the plugin with your preferred package manager:
-- Packer
use({
"piersolenski/wtf.nvim",
config = function()
require("wtf").setup()
end,
requires = {
"MunifTanjim/nui.nvim",
}
})
-- Lazy
{
"piersolenski/wtf.nvim",
dependencies = {
"MunifTanjim/nui.nvim",
},
opts = {},
keys = {
{
"gw",
mode = { "n", "x" },
function()
require("wtf").ai()
end,
desc = "Debug diagnostic with AI",
},
{
mode = { "n" },
"gW",
function()
require("wtf").search()
end,
desc = "Search diagnostic with Google",
},
},
}{
-- Default AI popup type
popup_type = "popup" | "horizontal" | "vertical",
-- An alternative way to set your API key
openai_api_key = "sk-xxxxxxxxxxxxxx",
-- ChatGPT Model
openai_model_id = "gpt-3.5-turbo",
-- Send code as well as diagnostics
context = true,
-- Set your preferred language for the response
language = "english",
-- Any additional instructions
additional_instructions = "Start the reply with 'OH HAI THERE'",
-- Default search engine, can be overridden by passing an option to WtfSeatch
search_engine = "google" | "duck_duck_go" | "stack_overflow" | "github",
-- Callbacks
hooks = {
request_started = nil,
request_finished = nil,
},
-- Add custom colours
winhighlight = "Normal:Normal,FloatBorder:FloatBorder",
}wtf.nvim works by sending the line's diagnostic messages along with contextual information (such as the offending code, file type and severity level) to various sources you can configure.
To use it, whenever you have an hint, warning or error in an LSP enabled environment, invoke one of the commands:
| Command | Modes | Description |
|---|---|---|
:Wtf [additional_instructions] |
Normal, Visual | Sends the diagnostic messages for a line or visual range to ChatGPT, along with the code if the context has been set to true. Additional instructions can also be specified, which might be useful if you want to refine the response further. |
:WtfSearch [search_engine] |
Normal | Uses a search engine (defaults to the one in the setup or Google if not provided) to search for the first diagnostic. It will attempt to filter out unrelated strings specific to your local environment, such as file paths, for broader results. |
You can add custom hooks to update your status line or other UI elements, for example, this code updates the status line colour to yellow whilst the request is in progress.
hooks = {
request_started = function()
vim.cmd("hi StatusLine ctermbg=NONE ctermfg=yellow")
end,
request_finished = vim.schedule_wrap(function()
vim.cmd("hi StatusLine ctermbg=NONE ctermfg=NONE")
end),
},There is a helper function get_status so that you can add a status component to lualine.
local wtf = require("wtf")
require('lualine').setup({
sections = {
lualine_x = { wtf.get_status },
}
})