thmsmlr / gpt.nvim

ChatGPT in your Neovim, customizable by you, for your usecases

gpt.nvim

---

Installation is easy. With your favorite package manager,

{
  "thmsmlr/gpt.nvim",
  config = function()
    require('gpt').setup({
      api_key = os.getenv("OPENAI_API_KEY")
    })

    opts = { silent = true, noremap = true }
    vim.keymap.set('v', '<C-g>r', require('gpt').replace, {
      silent = true,
      noremap = true,
      desc = "[G]pt [R]ewrite"
    })
    vim.keymap.set('v', '<C-g>p', require('gpt').visual_prompt, {
      silent = false,
      noremap = true,
      desc = "[G]pt [P]rompt"
    })
    vim.keymap.set('n', '<C-g>p', require('gpt').prompt, {
      silent = true,
      noremap = true,
      desc = "[G]pt [P]rompt"
    })
    vim.keymap.set('n', '<C-g>c', require('gpt').cancel, {
      silent = true,
      noremap = true,
      desc = "[G]pt [C]ancel"
    })
    vim.keymap.set('i', '<C-g>p', require('gpt').prompt, {
      silent = true,
      noremap = true,
      desc = "[G]pt [P]rompt"
    })
  end
}

You can get an API key via the OpenAI user settings page

It also requires that you have yq installed,

$ pacman -S yq

Usage

There are three ways to use the plugin:

1. In Visual Mode, select text and use it as the prompt with <C-g>p. ChatGPT will respond 2 lines below the selection.

explain-raw_EDIT_converted.mp4

2. In Insert Mode, <C-g>p will ask you for a prompt. ChatGPT will insert it's response into the buffer at the cursor's location.

write-haiku-raw_EDIT_converted.mp4

3. In Visual Mode, select text and rewrite it using <C-g>r. It'll ask you for the prompt to customize how GPT rewrites the selection.

edit-haiku-raw_EDIT_converted.mp4

gpt.nvim's Issues

Two small bugs

Hi, Interesting project!

It looks like this would be easier to extend than other projects, so I'm looking forward to playing with it.

I hope you don't mind a couple of bugs being reported. If you do mind, or are too busy no worries!

---

Bug1: Setup worked fine but I got the following error when calling require("gpt").prompt:

E5108: Error executing lua: /Users/<user>/.local/share/nvim/lazy/gpt.nvim/lua/gpt.lua:133: attempt to index local 'opts' (a nil value)
stack traceback:
        /Users/<user>/.local/share/nvim/lazy/gpt.nvim/lua/gpt.lua:133: in function 'create_response_writer'
        /Users/<user>/.local/share/nvim/lazy/gpt.nvim/lua/gpt.lua:241: in function </Users/<user>/.local/share/nvim/lazy/gpt.nvim/lua/gpt.lua:216>
E16: Invalid range

I was able to fix it by applying the following patch

diff --git a/lua/gpt.lua b/lua/gpt.lua
index 7bb1b0a..cd6feda 100644
--- a/lua/gpt.lua
+++ b/lua/gpt.lua
@@ -129,6 +129,7 @@ local function send_keys(keys)
 end
 
 local function create_response_writer(opts)
+    opts = opts or {}
 	local line_start = opts.line_no or vim.fn.line(".")
 	local bufnum = vim.api.nvim_get_current_buf()
 	local nsnum = vim.api.nvim_create_namespace("gpt")

---

Bug2: json_decode failed

I got the following error:

Vim:E474: Unexpected end of input: {
stack traceback:
^I[C]: in function 'json_decode'
^I/Users/<user>/.local/share/nvim/lazy/gpt.nvim/lua/gpt.lua:100: in function </Users/<user>/.local/share/nvim/lazy/gpt.nvim/lua/gpt.lua:92>

It looks like I'm getting the following back from OpenAI:

{
    "error": {
        "message": "The model: `gpt-4` does not exist",
        "type": "invalid_request_error",
        "param": null,
        "code": "model_not_found"
    }
}

I don't pay for the plus membership (yet), so it would be nice to be able to configure the model globally.
Also it might be nice to detect the error :)

---

Finally I see there are some undocumented functions like require("gpt").open_chatwindow 👀.
It looks like it would be interesting, so if you get the time to make a demo that would be very cool!

Thanks!