py/gh-commit
1
0
Fork 0
mirror of https://github.com/prdlk/gh-commit.git synced 2026-09-16 23:16:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat: rewrite gh-commit in Go with local Ollama backend

Browse Source 
Replaces the Python/Crush/DuckDB implementation with a CGO-free Go binary:
cobra CLI, modernc.org/sqlite storage, huh/lipgloss UI, and speed-tuned raw
/api/generate calls (think:false, keep_alive, capped num_predict) against a
local qwen3.5:2b. Ships as a gh extension via cli/gh-extension-precompile.
This commit is contained in:
t
2026-08-15 19:10:16 -04:00
parent 735229b150
commit 9647c854c9
31 changed files with 2731 additions and 1085 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+63 -67
View File
@@ -1,91 +1,87 @@
# gh-commit
AI-powered scoped git commits. Groups changes by project area, generates commit messages with `crush run`, and pushes — all in one command.
AI-powered scoped git commits as a GitHub CLI extension, driven by a local
Ollama model. Scopes map Conventional Commit scope names to path prefixes.
They are generated from the repository file tree, stored in a local SQLite
database, and auto-regenerate whenever `.gitignore` changes. Changed files are
grouped by scope and committed one scope at a time with generated
`type(scope): message` subjects.
## Install
```sh
# Prerequisites: git, a running Ollama server, and the model
ollama pull qwen3.5:2b
gh extension install prdlk/gh-commit
```
### Requirements
- [uv](https://docs.astral.sh/uv) — Python package runner (handles dependencies automatically)
- [Crush](https://github.com/charmbracelet/crush) — non-interactive AI runner
- `GROQ_API_KEY` — authenticates the default Groq model
By default, gh-commit uses Groq's `openai/gpt-oss-120b`. Override the model with `GH_COMMIT_CRUSH_MODEL`.
## Usage
```sh
# Initialize scopes for your repo (uses Crush to analyze structure)
gh commit init
# Commit changes grouped by scope
gh commit
# Auto-confirm + auto-push
gh commit --auto --push
# Manually refresh scopes after structural changes
gh commit refresh
# Sync scopes as GitHub labels
gh commit sync
cd your-repo
gh commit init # generate scopes for this repo
gh commit # commit changes grouped by scope
```
## How it works
1. **`gh commit init`** — `crush run` analyzes your repo structure and generates scope definitions (e.g., `core → src/`, `docs → docs/, README.md`, `ci → .github/workflows/`)
2. **`gh commit`** — Groups dirty files by scope, sends each staged diff to `crush run`, and commits each group separately
3. **Auto-refresh** — Whenever your `.gitignore` changes, scopes are automatically regenerated before committing (a content hash of `.gitignore` is tracked per repo)
4. Remaining unscoped files are handled in a final pass
5. Unpushed commits are offered for push
Scopes are stored in a local DuckDB database (`~/.local/share/gh-commit/gh-commit.db`) — no config files in your repo.
### Why Crush?
The former Mods roles now live directly in `smartcommit.py`, so gh-commit no longer depends on Mods configuration. Each generation invokes Crush's supported non-interactive mode with a self-contained prompt.
## Commands
| Command | Description |
|---------|-------------|
| `gh commit` | Commit changes grouped by scope |
| `gh commit init` | Generate scopes for current repo |
| `gh commit refresh` | Update scopes from current structure |
| `gh commit sync` | Sync scopes → GitHub labels |
| `gh commit list` | List all configured repositories |
| `gh commit remove` | Remove current repo from database |
| `gh commit db-path` | Print database file path |
| `gh commit version` | Print version |
| `gh commit help` | Show help |
| Command | Behavior |
|---|---|
| `gh commit` | Full scoped-commit flow: group changed files by scope, generate a message per group, confirm, commit, offer to push |
| `gh commit init` | Generate scopes from the file tree via Ollama; migrates legacy configs; confirms overwrite |
| `gh commit refresh` | Show current scopes, regenerate with existing scopes as context, confirm apply |
| `gh commit sync` | Create/update GitHub labels from scopes (color = first 6 hex chars of MD5 of the scope name) |
| `gh commit list` | Table of configured repositories: name, path, scope count |
| `gh commit remove` | Delete the current repo from the database after confirmation |
| `gh commit db-path` | Print the database file path |
| `gh commit version` | Print `gh-commit <version>` |
| `gh commit help` | Help, including the DB path and environment variable docs |
## Flags
| Flag | Description |
|------|-------------|
| `--auto` | Skip all confirmation prompts |
| `--push` | Auto-push after committing |
Flags: `--auto` (skip confirmations), `--push` (auto-push), `--model`, `--host`.
## Environment
| Variable | Description |
|----------|-------------|
| `GH_COMMIT_AUTO=1` | Skip all confirmation prompts |
| `GH_COMMIT_PUSH=1` | Auto-push after commits |
| `GH_COMMIT_NO_AUTO_REFRESH=1` | Don't auto-regenerate scopes when `.gitignore` changes |
| `GH_COMMIT_CRUSH_CMD` | Override the Crush command (default `crush`) |
| `GH_COMMIT_CRUSH_MODEL` | Override the Crush model (default `groq/openai/gpt-oss-120b`) |
| `GH_COMMIT_CRUSH_TIMEOUT` | Per-prompt timeout in seconds (default `120`) |
| `GH_COMMIT_DEBUG=1` | Show scope-response parse diagnostics |
| Variable | Default | Purpose |
|---|---|---|
| `GH_COMMIT_OLLAMA_HOST` | `http://localhost:11434` | Ollama server URL |
| `GH_COMMIT_MODEL` | `qwen3.5:2b` | Model tag |
| `GH_COMMIT_TIMEOUT` | `120` | Per-request timeout, seconds |
| `GH_COMMIT_AUTO` | `0` | `1` = skip all confirmations |
| `GH_COMMIT_PUSH` | `0` | `1` = auto-push after commits |
| `GH_COMMIT_NO_AUTO_REFRESH` | `0` | `1` = never auto-regenerate scopes |
| `GH_COMMIT_DEBUG` | `0` | `1` = print raw model output on parse failure |
## Migration
## Storage
Existing `.github/Repo.toml` or `.github/scopes.json` files are automatically detected and migrated to DuckDB on first run.
Scopes live in `${XDG_DATA_HOME:-~/.local/share}/gh-commit/gh-commit.sqlite`.
Legacy `.github/Repo.toml` and `.github/scopes.json` files are migrated
automatically on first run (the source file is archived as
`*.migrated.<timestamp>`). Databases from the old DuckDB-based Python version
are not migrated; rerun `gh commit init`.
## License
## Manual acceptance
MIT
```sh
mkdir /tmp/accept && cd /tmp/accept && git init
mkdir -p src docs
echo 'package app' > src/app.go
echo '# Docs' > docs/README.md
gh commit init # scopes generated and saved
echo '// change' >> src/app.go
echo 'More docs' >> docs/README.md
gh commit # one commit per scope
git log --format='%s' # verify type(scope): message subjects
```
## Development
```sh
go test ./... # unit tests
go test -tags integration -run TestSmoke . # needs a running Ollama server
go build .
```
Releases are built by `cli/gh-extension-precompile` on `v*` tags for
linux/amd64, linux/arm64, darwin/amd64, darwin/arm64, and windows/amd64
(`CGO_ENABLED=0`; the SQLite driver is pure Go).