gp - Git Push

Add, commit, and push in one command.

Generates commit messages using a configurable LLM with a heuristic fallback. Supports pushing multiple repos and waiting for all CI workflows in parallel.

Install

getpkg install gp

Usage

gp                              # Add all, auto-generate commit message, push
gp "Fix the parser"             # Custom commit message
gp --dry-run                    # Preview without executing
gp --staged-only                # Only commit staged changes
gp -y                           # Skip confirmations
gp -b develop                   # Push to a specific branch
gp --no-wait                    # Don't wait for CI after push
gp --llm ask                    # Use ask for commit messages
gp -y repo1/ repo2/ repo3/     # Push multiple repos, parallel CI wait

Options

Option	Description
-n, --dry-run	Preview changes without executing
-f, --force	Skip all confirmations
-y, --yes	Auto-answer yes to prompts
-a, --add-all	Add all files including untracked (default)
--staged-only	Only commit staged changes
-b, --branch	Push to specified branch
--no-wait	Skip waiting for CI workflow
--llm NAME	LLM for commit messages (see below)

Commit Message Generation

The LLM used for commit messages is configurable via --llm, GP_LLM env var, or git config gp.llm:

Value	Description
auto	Try ask, claude, ollama, then heuristic (default)
ask	Use the ask utility, configured by ~/.config/ask/ask.env
claude	Use Claude CLI (claude --print)
ollama	Use Ollama API with default model (qwen2.5:0.5b)
ollama:MODEL	Use Ollama API with a specific model (e.g. ollama:qwen2.5:1.5b)
none	Heuristic only, no LLM

ask reads its router settings from ~/.config/ask/ask.env, usually LLM_ROUTER_URL="http://rtx6000pro.home:8585". When ask is used, gp runs it with -v and reports the routed model when ask includes one, for example ask:qwen3.6-27b-fp8. Generated messages are shown with their source, for example Generated commit message using ask:qwen3.6-27b-fp8: 'Update gp commit message reporting'.

The Ollama API URL defaults to http://localhost:11434 and can be changed with GP_OLLAMA_URL or git config gp.ollama-url.

Heuristic fallback examples:

• Single file: Add src/auth.py, Update README.md, Remove old_config.json
• Multiple files: feat: Add 2 and update 3 files, docs: Update 2 files

Multi-Repo

Pass directory arguments to commit and push multiple repos, then wait for all CI workflows in parallel:

gp -y repo1/ repo2/ repo3/

Each repo gets its own auto-generated commit message. All commits and pushes happen first, then all CI workflows are monitored simultaneously. This replaces the old waterfall pattern of running gp in each repo one at a time.

CI Workflow Monitoring

After pushing to a Gitea remote, gp automatically monitors CI workflow progress:

[INFO] Waiting for CI workflows... (press ESC to stop)

  myproject
    ✓ build (linux/amd64)
    ⠹ build (linux/arm64) running
  otherproject
    ⠋ build (linux/amd64) running
    · build (linux/arm64) queued

• Press ESC at any time to stop waiting (CI continues in background)
• Requires a Gitea API token (from the remote URL, GITEA_TOKEN env, or git credential store)
• Automatically skipped for GitHub/GitLab remotes or when no token is available
• Use --no-wait to disable

Config Defaults

Set per-repo or global defaults via git config:

git config gp.auto-yes true           # Skip confirmations (like -y)
git config gp.wait false              # Don't wait for CI (like --no-wait)
git config gp.llm ask                 # Default LLM for commit messages
git config gp.ollama-url http://...   # Ollama API URL
git config --global gp.llm claude     # Apply globally