v3.1.0 - Copilot Extension Agents, Sticky Prompts, Reworked Contexts

[jellydn]

What's Changed

• Increase flexibility of diff selectors by @deathbeam in Increase flexibility of diff selectors #492
• Automatically register copilot-chat for treesitter markdown by @deathbeam in Automatically register copilot-chat for treesitter markdown #493
• Add support for copilot extension agents by @deathbeam in Add support for copilot extension agents #490
• Adjust how messages are built to match new vscode requests by @deathbeam in Adjust how messages are built to match new vscode requests #491
• Add agent descriptions to autocomplete by @deathbeam in Add agent descriptions to autocomplete #496
• Add support for #files (workspace file map) context by @deathbeam in Add support for #files (workspace file map) context #495
• style: improve model/agent selection prompts by @deathbeam in style: improve model/agent selection prompts #498
• feat(chat): improve auto follow cursor behavior by @taketwo in feat(chat): improve auto follow cursor behavior #497
• docs: add taketwo as a contributor for code by @allcontributors in docs: add taketwo as a contributor for code #499
• Remove private fields from chat class type definition by @deathbeam in Remove private fields from chat class type definition #500
• Add dotfyle badge by @deathbeam in Add dotfyle badge #501
• Add option to disable header highlighting by @deathbeam in Add option to disable header highlighting #502
• Add support for sticky prompt section with > by @deathbeam in Add support for sticky prompt section with >  #503
• Simplify how source buffer is being saved and unify prompt clearing by @deathbeam in Simplify how source buffer is being saved and unify prompt clearing #505
• feat(context): rewrite context system with input and resolving by @deathbeam in feat(context): rewrite context system with input and resolving #506
• docs: add missing help for model selection by @deathbeam in docs: add missing help for model selection #508
• docs: improve documentation for contexts and complete binding by @deathbeam in docs: improve documentation for contexts and complete binding #509
• Fix loading of default context by @deathbeam in Fix loading of default context #510
• Load chat auto_insert from config instead of passing it in contructor by @deathbeam in Load chat auto_insert from config instead of passing it in contructor #511
• feat: add automatic chat autocompletion by @deathbeam in feat: add automatic chat autocompletion #507
• Include filenames in buffer context selection, improve buffers context by @deathbeam in Include filenames in buffer context selection, improve buffers context #512
• Properly parse context inputs with : in them by @deathbeam in Properly parse context inputs with : in them #513
• refactor: improve semantic embeddings performance by @deathbeam in refactor: improve semantic embeddings performance #514
• perf: add caching for embeddings by @deathbeam in perf: add caching for embeddings #515
• docs: improve README with better examples and cleaner structure by @deathbeam in docs: improve README with better examples and cleaner structure #516
• Use relative paths for file references in contexts by @deathbeam in Use relative paths for file references in contexts #517
• Explicitely mention that multiple contexts are supported in README by @deathbeam in Explicitely mention that multiple contexts are supported in README #518
• Add documentation for selections to README by @deathbeam in Add documentation for selections to README #519
• Add table of contents to README because of all new config sections by @deathbeam in Add table of contents to README because of all new config sections #520
• Show errors inline in response by @deathbeam in Show errors inline in response #521
• refactor: improve diff handling implementation by @deathbeam in refactor: improve diff handling implementation #523

New Contributors

• @taketwo made their first contribution in feat(chat): improve auto follow cursor behavior #497

Full Changelog: v3.0.0...v3.1.0

---

This discussion was created from the release v3.1.0.

[jellydn]

Prompt Model selection

Models can now be selected directly in prompt with $ prefix.

Copilot Extension Agents

Agents are used to determine the AI agent used for the chat. You can list available agents with :CopilotChatAgents command.
You can set the agent in the prompt by using @ followed by the agent name.
Default "noop" agent is copilot.

For more information about extension agents, see here
You can install more agents from here

Sticky Prompts

You can set sticky prompt in chat by prefixing the text with > using markdown blockquote syntax.
The sticky prompt will be copied at start of every new prompt in chat window. You can freely edit the sticky prompt, only rule is > prefix at beginning of line.
This is useful for preserving stuff like context and agent selection (see below).
Example usage:

> #files

List all files in the workspace

> @models Using Mistral-small

What is 1 + 11

Reworked Contexts

Contexts are used to determine the context of the chat.
You can add context to the prompt by using # followed by the context name (multiple contexts are supported).
If context supports input, you can set the input in the prompt by using : followed by the input (or pressing complete key after :).
Default contexts are:

• buffer - Includes specified buffer in chat context (default current). Supports input.
• buffers - Includes all buffers in chat context (default listed). Supports input.
• file - Includes content of provided file in chat context. Supports input.
• files - Includes all non-hidden filenames in the current workspace in chat context. Supports input.
• git - Includes current git diff in chat context (default unstaged). Supports input.

You can define custom contexts very easily:

{
  contexts = {
    birthday = {
      input = function(callback)
        vim.ui.select({ 'user', 'napoleon' }, {
          prompt = 'Select birthday> ',
        }, callback)
      end,
      resolve = function(input)
        input = input or 'user'
        local birthday = input
        if input == 'user' then
          birthday = birthday .. ' birthday is April 1, 1990'
        elseif input == 'napoleon' then
          birthday = birthday .. ' birthday is August 15, 1769'
        end

        return {
          {
            content = birthday,
            filename = input .. '_birthday',
            filetype = 'text',
          }
        }
      end
    }
  }
}

Improved diffs

Diffs now support code excerpts on top of replacing whole selection when using /COPILOT_GENERATE.

Improved autocompletion

nvim-cmp autocompletion integration was removed in favour of custom simple autocomplete with tailored output specific to the plugin. This autocompletion works for everyone instead of relying on anything specific and can be disabled via chat_autocomplete property.

Completion results are also improved and they now contain also context, model and agent descriptions and with new context there is also manual trigger for inputs (e.g #context:).

Improved performance

Embedding processing was reworked and embeddings are now cached when used based on quick hash algorithm. This means that talking about same buffer/file or multiple files should be a lot faster.