diff and gitcommit cover commit buffers and diff output. cpp, kotlin and
nix were formatted by conform with no parser installed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Gitsigns was configured with sign glyphs and no bindings at all, so
hunks were visible but not navigable — the main thing you do when
reading back what an agent changed. Adds ]c/[c, preview, reset, blame
and diffthis, buffer-local on attach.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
root_dir was computed once, at startup, against whatever buffer happened
to be current — freezing one path for the session and breaking
lspconfig's local-binary lookup, which reads config.root_dir.
Everything overridden here is already handled better upstream: cmd
prefers the project-local biome, root_dir is monorepo-aware and resolved
per buffer, the filetypes list is wider, and the server only attaches
where a biome config applies. Only the formatting opt-out stays, since
conform owns that.
Also adds virtual_lines for the line under the cursor, which pairs with
the existing virtual_text = false.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Telescope ships in NvChad's own plugin set, so dropping it from neogit's
dependencies never removed it — it kept reinstalling. Disable it the same
way nvim-tree is disabled, and repoint the ten NvChad mappings that drove
it: five are already covered by fzf.lua bindings and are dropped, the
other five move to fzf-lua.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file required configs.lspconfig at spec-import time, which pulls in
nvchad.configs.lspconfig. On a fresh clone NvChad is not installed yet,
so the import threw and lazy skipped every spec below it — rustaceanvim,
typescript-tools, actions-preview, nui and lsp-endhints were all missing
until the second launch. Verified against a wiped profile.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The old one was the upstream starter's text: it told the reader to
delete .git and fork, and said nothing about this config.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
- inlay_hints = { enable = true } on lua_ls and solidity_ls: nothing in
nvim's LSP core reads that key. vim.lsp.inlay_hint.enable() already
turns them on.
- commented-out wgsl_analyzer block
- <leader>rn in on_attach: mappings.lua already binds it globally to the
same vim.lsp.buf.rename
- mini.files enabled = false: NvChad does not ship it, so there was
nothing to disable. The nvim-tree stub stays; that one is real.
- biome.json and .tool-versions: no JS in this repo
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The only path provider installed is cmp-async-path, which registers as
`async_path`. Our sources list overrode NvChad's index-by-index, so slot
5 became `path` — a name no source claims — and path completion silently
did nothing.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Agents run stylua/biome/rustfmt themselves, and a save-time reformat
races their in-flight writes to the same buffer. Timeout and LSP
fallback move to default_format_opts so <leader>fm (NvChad's binding)
keeps the same behaviour; conform now lazy-loads on first use.
Drops the /contrib/ skip — explicit invocation is already the opt-in.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Agent-driven work shifted nvim from writing code to reading and
reviewing it. Dropped:
- supermaven-nvim: inline AI completion duplicates the agent
- nvim-spectre: project-wide find/replace is now a prose request
- tailwind-tools: authoring aid for markup not hand-written
- editing.lua: hop.nvim plus ~85 lines of commented-out plugins
(hardtime, ufo, local-highlight)
- telescope.nvim: second picker; gd/gv/gr now use fzf-lua, which
was already bound to lsp_references
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The config uses the legacy `nvim-treesitter.configs` API, which the
`main` rewrite dropped. Without an explicit branch pin, `Lazy sync`
jumped to `main` and broke startup (module 'nvim-treesitter.configs'
not found). Pin both treesitter and treesitter-textobjects to `master`
and refresh the lockfile to a known-good synced state.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
naps62