Lima Language Server

Language Server Protocol support for Lima. The server reports parser errors and warnings, explains References 2.0 tokens on hover, and navigates ${document.path} references to their definitions. These features work in standalone .lima files and in Lima frontmatter in Markdown and MDX documents.

Usage

Configure an LSP client to run:

npx --yes @limaformat/lima-language-server --stdio

Use the language ID lima for .lima files. The server also accepts documents whose language ID is markdown or mdx and checks a leading --- frontmatter block. Whether an editor sends Markdown or MDX documents to this server is controlled by that editor's LSP client configuration, so embedded-frontmatter support is not available uniformly in every editor.

The setups below provide diagnostics and document-reference navigation. Sublime Text and Emacs can also use the syntax definitions shipped in this package. Standalone .lima files are the reliable target; attaching the server to Markdown or MDX is possible only when a client's document-routing configuration can send those documents with the matching language ID.

Editor setup

Neovim 0.11+

Neovim 0.11 added the built-in vim.lsp.config API. Add this to init.lua:

vim.filetype.add({
  extension = { lima = "lima" },
})

vim.lsp.config("lima_ls", {
  cmd = { "npx", "--yes", "@limaformat/lima-language-server", "--stdio" },
  filetypes = { "lima" },
})

vim.lsp.enable("lima_ls")

See Neovim's LSP configuration documentation. For older Neovim versions using the legacy nvim-lspconfig setup API, define the custom server explicitly:

vim.filetype.add({
  extension = { lima = "lima" },
})

local configs = require("lspconfig.configs")
local util = require("lspconfig.util")

if not configs.lima_ls then
  configs.lima_ls = {
    default_config = {
      cmd = { "npx", "--yes", "@limaformat/lima-language-server", "--stdio" },
      filetypes = { "lima" },
      root_dir = util.root_pattern(".git"),
      single_file_support = true,
    },
  }
end

require("lspconfig").lima_ls.setup({})

The legacy require("lspconfig").…setup() API is deprecated on Neovim 0.11 and newer; prefer the first configuration there.

Helix

Add this to your languages.toml:

[[language]]
name = "lima"
scope = "source.lima"
file-types = ["lima"]
language-id = "lima"
language-servers = ["lima-ls"]

[language-server.lima-ls]
command = "npx"
args = ["--yes", "@limaformat/lima-language-server", "--stdio"]

Helix has no built-in .lima association, so the [[language]] block is required, not optional.

Zed

The repository now includes a Zed extension that associates .lima files with this server. It is not published in Zed's extension registry yet; install it from a checkout with Extensions: Install Dev Extension and select the editors/zed directory. Zed then installs and updates the npm language server in the extension's private working directory.

The extension provides diagnostics and document-reference navigation. It uses tree-sitter-yaml only as the structural grammar needed to register the Lima language and does not claim accurate Lima syntax highlighting. See the extension README and Zed's development guide for the installation steps and current highlighting limitation.

Sublime Text

Copy syntaxes/lima.sublime-syntax to Packages/User/Lima.sublime-syntax in your Sublime Text data directory. The syntax assigns .lima files the source.lima scope and highlights Lima Core 1.0 plus active References 2.0 tokens.

Install Sublime's LSP package, then add this client under clients in Packages/User/LSP.sublime-settings:

{
  "clients": {
    "lima": {
      "enabled": true,
      "command": [
        "npx",
        "--yes",
        "@limaformat/lima-language-server",
        "--stdio"
      ],
      "selector": "source.lima"
    }
  }
}

The same entry can be placed in Sublime LSP's current custom-server file, Packages/User/LanguageServers.sublime-settings, without the surrounding clients object; see the client-configuration documentation.

The included syntax makes the source.lima selector usable for standalone .lima files. Markdown/MDX frontmatter injection is not included, so Sublime syntax highlighting is currently limited to standalone files.

JetBrains IDEs

The repository includes a JetBrains companion plugin that registers this server and maps *.lima automatically through LSP4IJ. Until it is published to JetBrains Marketplace, build and install its ZIP from disk as described in the companion README. No Generic LSP Server settings are needed after installing it.

For a manual setup without the companion, install LSP4IJ and add a user-defined (Generic LSP) server under Settings | Languages & Frameworks | Language Servers with:

Name: Lima
Command: npx --yes @limaformat/lima-language-server --stdio
Mapping: file name pattern *.lima
Language ID: lima

LSP4IJ's user-defined server guide shows the corresponding Server and Mappings tabs. This provides diagnostics and document-reference navigation. Neither setup installs Lima-specific syntax highlighting.

Emacs 29+

Add the directory containing emacs/lima-mode.el to load-path, then load the mode. It derives from prog-mode, associates the .lima extension, and provides Lima Core 1.0 and active References 2.0 font-lock highlighting. Eglot is built into Emacs 29; add its server entry and hook alongside the mode:

(add-to-list 'load-path "/path/to/lima/editors/lsp/emacs")
(require 'lima-mode)

(with-eval-after-load 'eglot
  (add-to-list 'eglot-server-programs
               '(lima-mode . ("npx"
                              "--yes"
                              "@limaformat/lima-language-server"
                              "--stdio"))))

(add-hook 'lima-mode-hook #'eglot-ensure)

See the GNU Eglot manual's server-setup documentation.

Configuration

The server accepts these settings through initializationOptions or workspace/didChangeConfiguration, nested under lima.diagnostics:

• lima.diagnostics.strict (default: true)
• lima.diagnostics.ignoreUnresolvedReferences (default: true)

Unresolved document references and partial-related findings are hidden by default. The shared setting is conservative: partial references need caller-provided context, while document references can be resolved locally.

Reference navigation

Hovering over an active ${key} or ${nested.path} token shows whether its document path resolves. Go to Definition jumps a resolved document reference to the line containing its target key. Lima's positioned parse tree records the target value's line but not the key's exact column, so the destination is the start of that line.

Hovering over $(key) explains that it is a partial reference whose value is supplied by the caller. Partial references intentionally have no definition target because Lima specifies no source-file convention for them. Quoted or commented reference-shaped text remains literal and receives neither feature.