agent-gitv1 0.1.4

AI-powered Git CLI using MCP + OpenAI, Gemini, or Ollama for commit workflows

agent-gitv1

AI-powered Git CLI for commit workflows using MCP plus your chosen LLM provider.

Supported providers:

• Google Gemini
• OpenAI
• Ollama (local or same-network)

Quick Start

1. Install:

pip install agent-gitv1

2. Configure provider/model:

agent config

3. Use it in any git repository:

agent commit
agent explain
agent push

Features

• Provider setup with agent config
• Smart staging with auto-grouping (UI/backend/config/etc.)
• Multi-suggestion commit messages (--suggestions)
• Repo-aware commit style learning (recent commits)
• agent explain for change summary, intent, and risk areas
• Ollama local-network detection (same /24, port 11434)
• Loading/thinking spinner during long-running tasks
• agent push with live command output streaming
• LLM-driven push-failure diagnosis with adaptive live git evidence collection

Requirements

• Python 3.10+
• Git
• uvx (pip install uv)
• Provider credentials (Gemini/OpenAI) or running Ollama server

Install

pip install agent-gitv1

Verify:

agent --version
agent --help

Configure

agent config

Provider notes:

• Gemini: set GEMINI_API_KEY or enter key in config
• OpenAI: set OPENAI_API_KEY or enter key in config
• Ollama: can auto-detect endpoints like http://192.168.1.35:11434

Commands

agent config

Configure LLM provider and model.

agent commit

Smart-stage changes, generate repo-aware suggestions, choose messages, and commit.

Examples:

agent commit
agent commit --suggestions 3
agent commit --no-smart-stage
agent commit --repo /path/to/repo
agent commit --verbose

agent explain

Explain current uncommitted changes:

• what changed
• why it likely changed
• risk areas
• files impacted

Examples:

agent explain
agent explain --repo /path/to/repo

agent push

Push to remote with live output.

Examples:

agent push
agent push --remote origin --branch main

When --branch is omitted, agent auto-detects and pushes your current local branch name.

If push fails, agent first uses the LLM to choose evidence depth, then runs diagnosis commands live. For divergence/history problems it collects full branch evidence (fetch, logs, ahead/behind, merge-base). For auth/permission issues it uses a lightweight evidence set to avoid noisy output. If the current branch has no upstream, agent offers to run git push --set-upstream automatically.

agent --help

Show all commands and options.

Environment Variables

• GEMINI_API_KEY
• OPENAI_API_KEY

Developer Setup (optional)

Only for contributors working on this codebase:

git clone <repo-url>
cd Agent_bhai
pip install -e .

Publish to PyPI

python -m pip install --upgrade build twine
python -m build
python -m twine check dist/*
python -m twine upload dist/*

License

MIT