LSP / Editor Integration

The LSP Server

The Unfault CLI includes a full Language Server Protocol server. Start it with:

unfault lsp

It communicates over stdio, which is the standard transport for most LSP clients.

The LSP server runs the same local analysis pipeline as unfault review (Tree-sitter parsing, semantic graph construction, and rule analysis) and publishes diagnostics in real time as you work.

What the LSP Provides

Diagnostics: Inline squiggles for findings (Error for critical/high, Warning for medium, Info for low)
Code actions: Quick-fix suggestions from structured patches or unified diffs
File centrality: Custom notification unfault/fileCentrality (how central the current file is in the dependency graph)
File dependencies: Custom notification unfault/fileDependencies (files that depend on the current one)
Hover: Function impact analysis using call graph traversal

Connecting an Editor

Neovim (nvim-lspconfig)

local lspconfig = require('lspconfig')
local configs = require('lspconfig.configs')

if not configs.unfault then
  configs.unfault = {
    default_config = {
      cmd = { 'unfault', 'lsp' },
      filetypes = { 'python', 'go', 'rust', 'typescript', 'javascript' },
      root_dir = lspconfig.util.root_pattern(
        'pyproject.toml', 'Cargo.toml', 'go.mod', 'package.json'
      ),
    },
  }
end

lspconfig.unfault.setup {}

VS Code (manual)

Create .vscode/settings.json with a custom LSP client extension, or use a generic LSP client extension that accepts a command. Point it at unfault lsp --stdio.

Supported Languages

Python
Go
Rust
TypeScript / JavaScript

Troubleshooting

No diagnostics appearing

Check the language server output: run unfault lsp --verbose manually and look for errors.
Verify unfault is on your PATH and runs without errors.
Check that your project root has a recognizable manifest file (pyproject.toml, Cargo.toml, go.mod, package.json, etc.).

Slow first analysis

The first run builds the full semantic graph for your project. Subsequent interactions use a cache.