🤖 Patch Pilot
📖 Live Documentation & Interactive FAQs: https://agent-docs-coral.vercel.app#patch-pilot
An autonomous test-driven code healing and bug patching agent with universal model support.
Patch Pilot inspects your local repository, analyzes bug descriptions, pinpoints root-cause files, iteratively applies patches, executes your test suite, and feeds back error stack traces until tests pass and changes are committed cleanly to Git.
🌐 Universal Model Support (Bring Your Own Key)
Patch Pilot works with all major AI providers or entirely local setups:
| Provider | Supported Models | Environment Variable |
|---|---|---|
| Anthropic Claude | claude-3-5-sonnet-20241022, claude-3-7-sonnet |
export ANTHROPIC_API_KEY="..." |
| OpenAI | gpt-4o, gpt-4o-mini, o1, o3-mini |
export OPENAI_API_KEY="..." |
| Google Gemini | gemini-2.5-flash, gemini-2.5-pro, gemini-1.5-pro |
export GEMINI_API_KEY="..." |
| DeepSeek | deepseek-chat, deepseek-reasoner |
export DEEPSEEK_API_KEY="..." |
| NVIDIA Nemotron | nvidia/nemotron-4-340b-instruct |
export NVIDIA_API_KEY="..." |
| OpenRouter | 200+ models across all frontier labs | export OPENROUTER_API_KEY="..." |
| Local / Self-Hosted | Ollama (qwen2.5-coder, llama3.2, deepseek-r1), vLLM |
--base-url http://localhost:11434/v1 |
🌟 Key Capabilities
- Multi-Model Intelligence: Fix code using Claude 3.5 Sonnet, GPT-4o, Gemini 2.5, DeepSeek-Coder, or local Ollama.
- Automatic Test Runner Detection: Supports
pytest,unittest,npm test,cargo test, andgo testwith zero configuration. - Iterative Self-Healing Loop: If a patch fails the test suite, Patch Pilot captures the exact error tracebacks and re-prompts the model to refine the fix.
- Git Branch & Commit Automation: Automatically isolates changes in a dedicated Git feature branch and writes conventional commit messages upon test verification.
🚀 Quick Start
1. Installation
git clone https://github.com/shubhransh-gupta/patch-pilot.git
cd patch-pilot
pip install -e .
⚡ Use as a GitHub Action in Your CI/CD
You can drop Patch Pilot directly into any repository's CI pipeline to automatically self-heal failing builds:
- name: Run Patch Pilot Self-Healing
uses: shubhransh-gupta/patch-pilot@v1
with:
api-key: ${{ secrets.GEMINI_API_KEY }}
provider: 'gemini'
issue: 'Fix failing unit tests in CI build'
2. Set Your Preferred API Key
Choose any provider:
# Option A: Anthropic Claude (Great for complex reasoning)
export ANTHROPIC_API_KEY="sk-ant-..."
# Option B: OpenAI
export OPENAI_API_KEY="sk-proj-..."
# Option C: Google Gemini
export GEMINI_API_KEY="AIzaSy..."
# Option D: DeepSeek
export DEEPSEEK_API_KEY="sk-..."
3. Run Patch Pilot Inside Any Repo
cd my-project/
# Auto-detects provider from your exported environment variable
patch-pilot "Fix KeyError when parsing optional config headers in client.py"
# Or explicitly choose provider and model
patch-pilot "Fix race condition in token refresher" --provider anthropic --model claude-3-5-sonnet-20241022
# Or run with local Ollama
patch-pilot "Fix off-by-one error in pagination" --base-url http://localhost:11434/v1 --model qwen2.5-coder
⚙️ Options
patch-pilot [ISSUE] [OPTIONS]
Options:
--provider, -p TEXT LLM provider: gemini, openai, anthropic, deepseek, nvidia, ollama, openrouter
--model, -m TEXT Specific model identifier
--base-url TEXT Custom API base URL (for Ollama, vLLM, etc.)
--repo, -r PATH Path to local repository (default: .)
--test-cmd, -t TEXT Custom test command (e.g. 'pytest tests/test_auth.py')
--max-attempts INT Max self-healing iterations (default: 4)
--api-key, -k TEXT Explicit API key override
--no-branch Skip automatic git branch creation
📦 Project Structure
patch-pilot/
├── patch_pilot/
│ ├── agent.py # Core self-healing loop and diagnosis orchestrator
│ ├── llm.py # Universal multi-provider LLM interface (Claude/GPT/Gemini/Ollama)
│ ├── runner.py # Subprocess test executor and framework detector
│ ├── git_ops.py # Git branch and commit wrapper
│ ├── config.py # Configuration and model settings
│ └── cli.py # Rich terminal interface
└── pyproject.toml # Packaging and entry points
📄 License
MIT License. Copyright (c) 2026 Shubhransh Gupta.
Metadata
Release files for patch-pilot-ai 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| patch_pilot_ai-0.1.0.tar.gz | 14.1 kB | Without provenance | Details for patch_pilot_ai-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| patch_pilot_ai-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for patch_pilot_ai-0.1.0-py3-none-any.whl |
Total release size: 28.2 kB
Release files / patch_pilot_ai-0.1.0.tar.gz
| Download URL | patch_pilot_ai-0.1.0.tar.gz |
|---|---|
| Size | 14.1 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
2d502a04e873caee6a3bdaa7671f24f7846f67a0c572c144d730c62c13edf811
|
|
BLAKE2b-256 checksum How to use checksums |
ef2e8f29f3938073147e24936136f7bcdb1bd273225217031179b2ec971e4c71
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.14.0
|
Release files / patch_pilot_ai-0.1.0-py3-none-any.whl
| Download URL | patch_pilot_ai-0.1.0-py3-none-any.whl |
|---|---|
| Size | 14.1 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
2b98b336dbba487e709a434f3e8fc17ed5613cb3922a7ef98bd65d0ebf5b20e9
|
|
BLAKE2b-256 checksum How to use checksums |
eef50ba76278f5df33cd3ff2008c31e3dca62d45fe7127359637a0d670c1e72b
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.14.0
|
Security for patch-pilot-ai 0.1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.