Neovim Configuration

This is my Neovim setup largely based and expanded on Josean Martinez's blog post and the youtube video

Other than small retouch the setup is virtually identitcal to Josean's setup, checkout his video for more detailed explanation on the setup.

Requirements

• Neovim 0.12 or later.
• A C compiler on PATH (e.g. gcc from MinGW) to build treesitter parsers and telescope-fzf-native. Without MSVC, the config points tree-sitter at gcc by setting CC. nvim-treesitter also needs the tree-sitter CLI, which Mason installs; on a fresh machine parsers install once Mason has it.
• For Rust, rust-analyzer from your toolchain: rustup component add rust-analyzer. rustaceanvim starts it (and Mason's codelldb for debugging); don't install rust-analyzer through Mason.
• For octo.nvim (GitHub issues and PRs, <leader>o…), the GitHub CLI, logged in with gh auth login.
• For C#, the .NET 10 SDK, which the pinned Roslyn needs to install and run. Mason installs Roslyn (the language server VS Code's C# extension uses) when dotnet is on PATH, and netcoredbg. <leader>dc in a C# file offers Build and debug a project: pick a runnable project from the solution, and it builds it (build errors go to quickfix) and starts the debugger on its output.

WSL

The same config runs in WSL (Ubuntu 24.04):

• Neovim comes from Homebrew (brew install neovim), which ~/.bashrc puts on PATH with eval "$(/home/linuxbrew/.linuxbrew/bin/brew shellenv bash)". Upgrade it with brew upgrade neovim; CI tests the release pinned in .github/actions/setup-nvim/action.yml.
• Config: a clone of this repo in ~/.config/nvim. Update it with git -C ~/.config/nvim pull, then run :Lazy restore to check out the plugin commits in lazy-lock.json. Mason installs the language servers on the first interactive start (mason-lspconfig skips them in headless runs). Plugins, Mason packages and parsers live in ~/.local/share/nvim; to reinstall them from scratch, move its lazy, mason and site folders aside and restart.
• Tools: build-essential, unzip, ripgrep and fd-find from apt (with fdfind linked as ~/.local/bin/fd), Node.js from nvm for Mason's npm packages, Go, rustup with rust-analyzer, lazygit (brew install lazygit) and the GitHub CLI (brew install gh). Without dotnet, Mason skips Roslyn, so C# has no language server.
• Clipboard: unnamedplus goes through win32yank.exe from the Windows Neovim install, which WSL finds on the Windows PATH.

Updates

• Plugins: .github/workflows/plugin-update.yml runs :Lazy update every Monday (or on demand from the Actions tab), commits the new lazy-lock.json to simonk/dev/plugin-update, and opens or refreshes an Update plugins PR that links each plugin's changes. CI runs on that PR, so the smoke test covers the new commits; merge it if it's green. The workflow force-pushes the branch each week, but leaves it alone once it has a commit from anyone else, so you can push a fix to it. It needs Allow GitHub Actions to create and approve pull requests turned on in Settings → Actions → General.
• GitHub Actions: Dependabot (.github/dependabot.yml) opens one grouped PR a month for the actions used in workflows and .github/actions/. It skips generated *.lock.yml workflows, because gh aw compile would overwrite its edits; refresh those with gh aw upgrade. Dependabot can't update lazy.nvim plugins, hence the workflow above.

CI checks

Every pull request (and push to main) runs .github/workflows/ci.yml:

• StyLua: stylua --check . using .stylua.toml. Fix locally with stylua ..
• Startup smoke test (Windows): installs every plugin from lazy-lock.json and every treesitter parser into empty directories, opens a Lua and a C# buffer, loads all plugins, and fails on any error. It also checks wiring that wouldn't raise an error: every mason-lspconfig server name resolves to an LSP config, Go/TSX/Lua/Rust buffers get the expected formatter and treesitter highlighting, files without a formatter get trailing whitespace trimmed (patches excepted), and web filetypes use 2-space indents. Mason package download failures are ignored because they depend on the network, not the config; language servers aren't installed in headless runs, so they're never started.

Run the smoke test locally from this directory:

nvim --headless --cmd "luafile tests/smoke.lua"

In WSL, run the same command from ~/.config/nvim.

To hide the bulk reformat commit from git blame, run git config blame.ignoreRevsFile .git-blame-ignore-revs.

AI PR review

pr-review-command.md in .github/workflows/ is an AI reviewer written as GitHub Agentic Workflows. It only runs when you comment /review on a PR (drafts included), so it never uses premium requests on its own. It only comments; it never approves or blocks.

• How it works (shared/pr-review.md): two reviewers on different models (claude-reviewer, gpt-reviewer) review the PR independently. A lead agent merges their findings, asks each reviewer to check what only the other found, verifies disputes against the code, and posts the result with a Review context and counts table. Consensus reached: False or a low confidence score means the round deserves a second look; comment /review to retry.
• The review criteria live in .github/agents/pr-reviewer.agent.md, which is also available in the repo's Agents tab. The sub-agents' task and output format are in shared/pr-reviewer-subagent.md.
• Models are pinned: the lead's in pr-review-command.md, the reviewers' in the ## agent: blocks of shared/pr-review.md.
• The agent always uses the instructions on main, not the PR's copy, so a PR can't rewrite its own reviewer. A PR that changes the reviewer is still reviewed by the old version; test the change with /review after merging.
• Edit the .md files, then run gh aw compile (needs gh extension install github/gh-aw) to regenerate the .lock.yml files GitHub actually runs.
• It runs Copilot with the COPILOT_GITHUB_TOKEN repository secret: a fine-grained personal access token with the Copilot Requests permission. Each run uses Copilot premium requests for three models (the lead and two reviewers). Only users with write access trigger it.