Skip to content

mugiwaraluffy56/gitdaemon

BranchesTags

Repository files navigation

gitdaemon (gd)

Declarative background Git sync engine — auto-stage, commit, push, and rebase while you work

CI Crates.io License: MIT MSRV

gd is a daemon that runs next to your editor and handles the entire Git loop — watching for file changes, staging, generating meaningful commit messages, pushing, and keeping your branch rebased on the latest base — so you never have to stop and run git add / commit / push / rebase manually.

Features

Feature Details
Heuristic commit messages Generates conventional commits (feat, fix, refactor, chore, docs, test, build) by parsing symbol declarations from the actual diff — no API key needed
AI commit messages Optional Claude integration produces higher-quality messages; falls back to the heuristic automatically
Auto-stage Watches the working tree (150 ms debounce), respects .gitignore and your custom ignore list
Batched push Queues commits and pushes in intervals — no hammering the remote
Background fetch Keeps local refs fresh without running git fetch
Auto-sync base branch Fast-forwards main and rebases your feature branch onto it with auto-stash
Secret scanning Scans diffs for API keys, tokens, and private keys before every push
Branch-push guard Blocks auto-push to protected branches (main, master, develop)
Commit strategies time (every N seconds) or change_count (every N files)
Hooks pre_commit, post_commit, on_push_success, on_conflict shell hooks
IPC control gd pause, gd push, gd status, gd undo, gd squash, gd ls via Unix socket
Conflict guard Detects merge conflicts, pauses push, fires on_conflict hook

Installation

Pre-built binary

# macOS (Apple Silicon)
curl -L https://github.com/mugiwaraluffy56/gitdaemon/releases/latest/download/gd-aarch64-apple-darwin.tar.gz \
  | tar xz && sudo mv gd /usr/local/bin/

# macOS (Intel)
curl -L https://github.com/mugiwaraluffy56/gitdaemon/releases/latest/download/gd-x86_64-apple-darwin.tar.gz \
  | tar xz && sudo mv gd /usr/local/bin/

# Linux (x86_64)
curl -L https://github.com/mugiwaraluffy56/gitdaemon/releases/latest/download/gd-x86_64-unknown-linux-gnu.tar.gz \
  | tar xz && sudo mv gd /usr/local/bin/

From source

git clone https://github.com/mugiwaraluffy56/gitdaemon
cd gitdaemon
cargo build --release
cp target/release/gd ~/.local/bin/

See docs/installation.md for shell completions, updates, and uninstall.

Quick start

cd ~/projects/my-project   # any git repo

gd init                    # create gd.yml
gd up -d                   # start daemon in the background

# Work normally — gd auto-stages, commits, and pushes

gd status                  # check what's happening
gd ls                      # see which files are being tracked
gd log                     # view recent auto-commits
gd down                    # stop the daemon

Commit messages

gd generates conventional commit messages directly from the diff, not just from file names. Two modes are available and can be switched in gd.yml:

Heuristic (default — no API key needed)

The built-in generator parses symbol declarations (struct, enum, fn, class, def, etc.) from diff lines using language-aware regexes:

feat(git): introduce PushQueue and implement try_push
fix(config): rework validate to reject zero-value intervals
refactor(daemon): extract IPC server into dedicated module
chore: drop stale lock files
test(secrets): add coverage for AWS key and GitHub token patterns
docs: update README with AI commit messages section
build: add reqwest and dotenvy dependencies

Supported languages: Rust, TypeScript, JavaScript, Python, Go.

AI-generated (optional — any LLM)

When ai.enabled: true, gd sends the staged diff to any LLM and uses the response as the commit message. Supports Anthropic Claude, OpenAI, Groq, Together AI, Ollama, LM Studio, and any OpenAI-compatible endpoint. Falls back to the heuristic automatically on any error. Commits are never blocked by the AI path.

# Anthropic Claude
ai:
  enabled: true
  provider: anthropic
  api_key: ""                           # reads ANTHROPIC_API_KEY from env/.env
  model: "claude-haiku-4-5-20251001"

# OpenAI
ai:
  enabled: true
  provider: openai
  api_key: ""                           # reads OPENAI_API_KEY from env/.env
  model: "gpt-4o-mini"

# Ollama (local, no key needed)
ai:
  enabled: true
  provider: openai
  api_key: "local"
  model: "llama3"
  base_url: "http://localhost:11434"

See docs/ai-commit-messages.md for all providers, model examples, privacy considerations, and troubleshooting.

Customising the template

commit:
  message: "{summary}"   # default
  # message: "[{branch}] {summary} ({count} files, {timestamp})"

Tokens: {summary} · {count} · {branch} · {timestamp}

Configuration (gd.yml)

version: 1

repo:
  auto_stage: true        # stage changes automatically
  auto_fetch: true        # fetch from remotes in the background
  fetch_interval: 60      # seconds between fetches
  auto_sync_base: true    # fast-forward main and rebase your branch
  base_branch: main
  rebase_on_sync: true

commit:
  strategy: time          # "time" or "change_count"
  interval: 120           # seconds between auto-commits
  change_threshold: 10    # files before committing (change_count strategy)
  message: "{summary}"

push:
  strategy: batch
  interval: 300
  branch: main
  protected_branches: [main, master, develop]

ignore:
  - "*.log"
  - "node_modules/"
  - ".env"

safety:
  confirm_first: false
  block_secrets: true     # scan diffs for secrets before every push

hooks:
  pre_commit: ""          # e.g. "cargo fmt --check"
  post_commit: ""         # e.g. "cargo test --quiet"
  on_push_success: ""     # env: $FG_BRANCH, $FG_COMMITS
  on_conflict: ""         # env: $FG_BRANCH, $FG_ERROR

ai:
  enabled: false          # set true to enable AI commit messages
  provider: anthropic     # "anthropic" or "openai" (covers Groq, Ollama, etc.)
  api_key: ""             # blank = auto-read ANTHROPIC_API_KEY / OPENAI_API_KEY
  model: ""               # required — e.g. claude-haiku-4-5-20251001, gpt-4o-mini
  base_url: ""            # leave blank for default; set for local/custom endpoints
  max_diff_chars: 12000

See examples/gd.full.yml for a fully-annotated config and docs/config-reference.md for the complete field reference.

CLI reference

Daemon lifecycle

gd up            # start daemon (foreground, Ctrl-C to stop)
gd up -d         # start daemon detached in the background
gd down          # send SIGTERM to the running daemon

Status & history

gd status        # live sync state: branch, ahead/behind, staged, health
gd ls            # list tracked files by state (staged/modified/untracked/…)
gd log           # last 10 gd auto-commits
gd log -n 25     # last 25
gd log -f        # follow mode — stream new commits in real time
gd log --all     # all commits, not just gd ones

History editing

gd undo          # soft-reset last auto-commit back to index
gd undo 3        # soft-reset last 3 auto-commits
gd undo --force  # undo even non-gd commits
gd squash 5      # squash last 5 auto-commits into one

Control

gd pause         # pause auto-push (staging + committing continue)
gd resume        # resume auto-push
gd push          # trigger immediate push
gd init          # create gd.yml in current repo
gd init --force  # overwrite existing gd.yml

gd status output

⚡ gitdaemon — /home/alice/projects/my-project
  branch   feature/auth → origin/feature/auth
  ahead    3 commits (queued to push, last push 42s ago)
  behind   0
  staged   2 files
  watching 1,204 files
  daemon   running (pid 12345)

Health icons: ⚡ healthy · ⬇ behind remote · ⏸ push paused · ✗ error

Security

  • Secret scanning before every push — AWS keys, GitHub tokens, Stripe keys, Google API keys, Slack tokens, private key headers, and generic credential patterns. See docs/secret-scanning.md.
  • Protected branch guard — auto-push blocked on main, master, develop.
  • AI privacy — when enabled, the staged diff is sent to Anthropic's API. See docs/ai-commit-messages.md#privacy-considerations.

Documentation

Document Description
docs/installation.md Install, update, uninstall, shell completions
docs/user-guide.md Full feature walkthrough
docs/config-reference.md Every gd.yml field documented
docs/ai-commit-messages.md Claude integration setup and guide
docs/hooks.md Shell hook reference with real-world examples
docs/secret-scanning.md Scanner patterns and false-positive handling
docs/architecture.md Internal design and data flow diagrams
docs/troubleshooting.md Common problems and fixes
CHANGELOG.md Version history
CONTRIBUTING.md How to contribute

Development

cargo build              # debug build
cargo build --release    # optimised build
cargo check              # fast type + borrow check
cargo clippy             # lint
cargo fmt                # format
cargo test               # all tests
cargo bench              # benchmarks → target/criterion/
RUST_LOG=debug gd up     # daemon with debug logging

Use just for common workflows:

just ci       # fmt-check + clippy + test
just bench    # run benchmarks
just docs     # cargo doc --open
just install  # cargo install --path .

Non-goals

  • Not a replacement for intentional, context-rich commits during code review
  • Does not resolve merge conflicts automatically
  • Not designed for monorepos where CI gates every push
  • Does not support Windows (Unix sockets required for IPC)

Contributing

See CONTRIBUTING.md. Bug reports and feature requests go through GitHub Issues.

License

MIT — Copyright © 2022 gitdaemon contributors

About

Declarative background Git sync engine with intelligent auto commits

Resources

Readme

License

MIT license

Code of conduct

Code of conduct

Contributing

Contributing

Security policy

Security policy
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages