Awesome

cmp-ai

AI source for hrsh7th/nvim-cmp

This is a general purpose AI source for cmp, easily adapted to any restapi supporting remote code completion.

For now, HuggingFace, SantaCoder, OpenAI Chat, Codestral, Ollama and Google Bard are implemented.

Install

Dependencies

• You will need plenary.nvim to use this plugin.
• For using Codestral, OpenAI or HuggingFace, you will also need curl.
• For using Google Bard, you will need dsdanielpark/Bard-API.

Using a plugin manager

Using Lazy:

return require("lazy").setup({
    {'tzachar/cmp-ai', dependencies = 'nvim-lua/plenary.nvim'},
    {'hrsh7th/nvim-cmp', dependencies = {'tzachar/cmp-ai'}},
})

And later, tell cmp to use this plugin:

require'cmp'.setup {
    sources = {
        { name = 'cmp_ai' },
    },
}

Setup

Please note the use of : instead of a .

To use HuggingFace:

local cmp_ai = require('cmp_ai.config')

cmp_ai:setup({
  max_lines = 1000,
  provider = 'HF',
  notify = true,
  notify_callback = function(msg)
    vim.notify(msg)
  end,
  run_on_every_keystroke = true,
  ignored_file_types = {
    -- default is not to ignore
    -- uncomment to ignore in lua:
    -- lua = true
  },
})

You will also need to make sure you have the Hugging Face api key in you environment, HF_API_KEY.

To use OpenAI:

local cmp_ai = require('cmp_ai.config')

cmp_ai:setup({
  max_lines = 1000,
  provider = 'OpenAI',
  provider_options = {
    model = 'gpt-4',
  },
  notify = true,
  notify_callback = function(msg)
    vim.notify(msg)
  end,
  run_on_every_keystroke = true,
  ignored_file_types = {
    -- default is not to ignore
    -- uncomment to ignore in lua:
    -- lua = true
  },
})

You will also need to make sure you have the OpenAI api key in you environment, OPENAI_API_KEY.

Available models for OpenAI are gpt-4 and gpt-3.5-turbo.

To use Codestral:

local cmp_ai = require('cmp_ai.config')

cmp_ai:setup({
  max_lines = 1000,
  provider = 'Codestral',
  provider_options = {
    model = 'codestral-latest',
  },
  notify = true,
  notify_callback = function(msg)
    vim.notify(msg)
  end,
  run_on_every_keystroke = true,
  ignored_file_types = {
    -- default is not to ignore
    -- uncomment to ignore in lua:
    -- lua = true
  },
})

You will also need to make sure you have the Codestral api key in you environment, CODESTRAL_API_KEY.

You can also use the suffix and prompt parameters, see Codestral for more details.

local cmp_ai = require('cmp_ai.config')

cmp_ai:setup({
  max_lines = 1000,
  provider = 'Codestral',
  provider_options = {
    model = 'codestral-latest',
    prompt = function(lines_before, lines_after)
      return lines_before
    end,
    suffix = function(lines_after)
      return lines_after
    end
  },
  notify = true,
  notify_callback = function(msg)
    vim.notify(msg)
  end,
  run_on_every_keystroke = true,
})

To use Google Bard:

local cmp_ai = require('cmp_ai.config')

cmp_ai:setup({
  max_lines = 1000,
  provider = 'Bard',
  notify = true,
  notify_callback = function(msg)
    vim.notify(msg)
  end,
  run_on_every_keystroke = true,
  ignored_file_types = {
    -- default is not to ignore
    -- uncomment to ignore in lua:
    -- lua = true
  },
})

You will also need to follow the instructions on dsdanielpark/Bard-API to get the __Secure-1PSID key, and set the environment variable BARD_API_KEY accordingly (note that this plugin expects BARD_API_KEY without a leading underscore).

To use Ollama:

local cmp_ai = require('cmp_ai.config')

cmp_ai:setup({
  max_lines = 100,
  provider = 'Ollama',
  provider_options = {
    model = 'codellama:7b-code',
  },
  notify = true,
  notify_callback = function(msg)
    vim.notify(msg)
  end,
  run_on_every_keystroke = true,
  ignored_file_types = {
    -- default is not to ignore
    -- uncomment to ignore in lua:
    -- lua = true
  },
})

With Ollama you can also use the suffix parameter, typically when you want to use cmp-ai for code completion and you want to use the default plugin/prompt.

If the model you're using has the following template:

{{- if .Suffix }}<|fim_prefix|>{{ .Prompt }}<|fim_suffix|>{{ .Suffix }}<|fim_middle|>
{{- else }}{{ .Prompt }}
{{- end }}

then you can use the suffix parameter to not change the prompt. since the model will use your suffix and the prompt to construct the template. The prompts should be the lines_before and suffix the lines_after Now you can even change the model without the need to adjust the prompt or suffix functions.

local cmp_ai = require('cmp_ai.config')

cmp_ai:setup({
  max_lines = 100,
  provider = 'Ollama',
  provider_options = {
    model = 'codegemma:2b-code',
    prompt = function(lines_before, lines_after)
      return lines_before
    end,
    suffix = function(lines_after)
      return lines_after
    end,
  },
  notify = true,
  notify_callback = function(msg)
    vim.notify(msg)
  end,
  run_on_every_keystroke = true,
})

[!NOTE] Different models may implement different special tokens to delimit prefix and suffix. You may want to consult the official documentation for the specific tokens used for your model and the recommended format of the prompt. For example, qwen2.5-coder used <|fim_prefix|>, <|fim_middle|> and <|fim_suffix|> (as well as some other special tokens for project context) as the delimiter for fill-in-middle code completion and provided examples on how to construct the prompt. This is model-specific and Ollama supports all kinds of different models and fine-tunes, so it's best if you write your own prompt like the following example:

local cmp_ai = require('cmp_ai.config')

cmp_ai:setup({
  max_lines = 100,
  provider = 'Ollama',
  provider_options = {
    model = 'qwen2.5-coder:7b-base-q6_K',
    prompt = function(lines_before, lines_after)
    -- You may include filetype and/or other project-wise context in this string as well.
    -- Consult model documentation in case there are special tokens for this.
      return "<|fim_prefix|>" .. lines_before .. "<|fim_suffix|>" .. lines_after .. "<|fim_middle|>"
    end,
  },
  notify = true,
  notify_callback = function(msg)
    vim.notify(msg)
  end,
  run_on_every_keystroke = false,
})

[!NOTE] It's also worth noting that, for some models (like qwen2.5-coder), the base model appears to be better for completion because it only replies with the code, whereas the instruction-tuned variant tends to reply with a piece of Markdown text which cannot be directly used as the completion candidate.

To use Tabby:

local cmp_ai = require('cmp_ai.config')

cmp_ai:setup({
  max_lines = 1000,
  provider = 'Tabby',
  notify = true,
  provider_options = {
    -- These are optional
    -- user = 'yourusername',
    -- temperature = 0.2,
    -- seed = 'randomstring',
  },
  notify_callback = function(msg)
    vim.notify(msg)
  end,
  run_on_every_keystroke = true,
  ignored_file_types = {
    -- default is not to ignore
    -- uncomment to ignore in lua:
    -- lua = true
  },
})

You will also need to make sure you have the Tabby api key in your environment, TABBY_API_KEY.

notify

As some completion sources can be quit slow, setting this to true will trigger a notification when a completion starts and ends using vim.notify.

notify_callback

The default notify function uses vim.notify, but an override can be configured. For example:

notify_callback = function(msg)
  require('notify').notify(msg, vim.log.levels.INFO, {
    title = 'OpenAI',
    render = 'compact',
  })
end

max_lines

How many lines of buffer context to use

run_on_every_keystroke

Generate new completion items on every keystroke.

ignored_file_types (table: <string:bool>)

Which file types to ignore. For example:

local ignored_file_types = {
  html = true,
}

cmp-ai will not offer completions when vim.bo.filetype is html.

Dedicated cmp keybindings

As completions can take time, and you might not want to trigger expensive apis on every keystroke, you can configure cmp-ai to trigger only with a specific key press. For example, to bind cmp-ai to <c-x>, you can do the following:

cmp.setup({
  ...
  mapping = {
    ...
    ['<C-x>'] = cmp.mapping(
      cmp.mapping.complete({
        config = {
          sources = cmp.config.sources({
            { name = 'cmp_ai' },
          }),
        },
      }),
      { 'i' }
    ),
  },
})

Also, make sure you do not pass cmp-ai to the default list of cmp sources.

Pretty Printing Menu Items

You can use the following to pretty print the completion menu (requires lspkind and patched fonts (https://www.nerdfonts.com)):

require('cmp').setup({
  sources = {
    { name = 'cmp_ai' },
  },
  formatting = {
    format = require('lspkind').cmp_format({
      mode = "symbol_text",
      maxwidth = 50,
      ellipsis_char = '...',
      show_labelDetails = true,
      symbol_map = {
        HF = "",
        OpenAI = "",
        Codestral = "",
        Bard = "",
      }
    });
  },
})

Sorting

You can bump cmp-ai completions to the top of your completion menu like so:

local compare = require('cmp.config.compare')
cmp.setup({
  sorting = {
    priority_weight = 2,
    comparators = {
      require('cmp_ai.compare'),
      compare.offset,
      compare.exact,
      compare.score,
      compare.recently_used,
      compare.kind,
      compare.sort_text,
      compare.length,
      compare.order,
    },
  },
})