GitBot 🤖

A powerful GitLab bot that provides Copilot-like LLM assistance for GitLab repositories

GitBot monitors GitLab repositories for mentions and responds with AI-powered summaries and assistance for issues and merge requests. It provides intelligent code analysis, repository context, and automated responses without requiring GitLab Ultimate licensing.

✨ Features

• 🔍 Smart Repository Analysis: Automatically indexes and searches repository content for relevant context
• 🤖 AI-Powered Responses: Uses OpenAI's GPT models to provide intelligent summaries and assistance
• 📝 Issue & MR Support: Responds to mentions in both issues and merge requests with contextual information
• 🏷️ Stale Issue Management: Automatically tracks and labels stale issues based on configurable time periods
• 🔄 Real-time Polling: Continuously monitors repositories for new activity and mentions
• 🐳 Docker Ready: Fully containerized for easy deployment and scaling
• ⚡ High Performance: Built with Rust for optimal speed and memory efficiency

🚀 Quick Start

Prerequisites

• Rust (latest stable version recommended)
• GitLab account with API access
• OpenAI API key

Installation

1. Clone the repository:

   git clone https://github.com/myaple/gitbot.git
   cd gitbot

2. Build the project:

   cargo build --release

3. Run with configuration:

   ./target/release/gitbot \
     --gitlab-token YOUR_GITLAB_TOKEN \
     --openai-api-key YOUR_OPENAI_KEY \
     --bot-username YOUR_BOT_USERNAME \
     --repos-to-poll group/project1,group/project2 \
     --context-lines 15

Using Environment Variables

export GITBOT_GITLAB_TOKEN=YOUR_GITLAB_TOKEN
export GITBOT_OPENAI_API_KEY=YOUR_OPENAI_KEY  
export GITBOT_BOT_USERNAME=YOUR_BOT_USERNAME
export GITBOT_REPOS_TO_POLL=group/project1,group/project2
export GITBOT_CONTEXT_LINES=15
./target/release/gitbot

Docker Deployment

1. Build the Docker image:

   docker build -t gitbot .

2. Run the container:

   docker run -d --name gitbot \
     -e GITBOT_GITLAB_TOKEN="YOUR_GITLAB_TOKEN" \
     -e GITBOT_OPENAI_API_KEY="YOUR_OPENAI_KEY" \
     -e GITBOT_BOT_USERNAME="YOUR_BOT_USERNAME" \
     -e GITBOT_REPOS_TO_POLL="group/project1,group/project2" \
     -e GITBOT_CONTEXT_LINES="15" \
     gitbot

⚙️ Configuration

GitBot supports extensive configuration through command line arguments or environment variables:

Environment Variable	CLI Argument	Default	Description
GITBOT_GITLAB_URL	--gitlab-url	https://gitlab.com	GitLab instance URL
GITBOT_GITLAB_TOKEN	--gitlab-token	-	GitLab API token (required)
GITBOT_OPENAI_API_KEY	--openai-api-key	-	OpenAI API key (required)
GITBOT_BOT_USERNAME	--bot-username	-	Bot username on GitLab (required)
GITBOT_REPOS_TO_POLL	--repos-to-poll	-	Comma-separated list of repositories (required)
GITBOT_OPENAI_MODEL	--openai-model	gpt-3.5-turbo	OpenAI model to use
GITBOT_POLL_INTERVAL_SECONDS	--poll-interval-seconds	60	Polling interval in seconds
GITBOT_STALE_ISSUE_DAYS	--stale-issue-days	30	Days after which issues are marked stale
GITBOT_CONTEXT_LINES	--context-lines	10	Number of lines before and after keyword matches
GITBOT_LOG_LEVEL	--log-level	info	Log level (trace, debug, info, warn, error)

View all configuration options

Environment Variable	CLI Argument	Default	Description
GITBOT_OPENAI_CUSTOM_URL	--openai-custom-url	https://api.openai.com/v1	Custom OpenAI API URL
GITBOT_OPENAI_TEMPERATURE	--openai-temperature	0.7	Temperature for AI responses (0.0-1.0)
GITBOT_OPENAI_MAX_TOKENS	--openai-max-tokens	1024	Maximum tokens in AI responses
GITBOT_MAX_AGE_HOURS	--max-age-hours	24	Maximum age for issues/MRs to process
GITBOT_CONTEXT_REPO_PATH	--context-repo-path	-	Additional repository for context
GITBOT_MAX_CONTEXT_SIZE	--max-context-size	60000	Maximum context characters
GITBOT_DEFAULT_BRANCH	--default-branch	main	Default branch name
GITBOT_PROMPT_PREFIX	--prompt-prefix	-	Optional prefix to prepend to every prompt sent to the LLM
GITBOT_CLIENT_CERT_PATH	--client-cert-path	-	Path to client certificate file for OpenAI API
GITBOT_CLIENT_KEY_PATH	--client-key-path	-	Path to client private key file for OpenAI API
GITBOT_CLIENT_KEY_PASSWORD	env only	-	Password for client private key (environment variable only)

📖 Usage

1. Set up the bot: Configure GitBot with your GitLab and OpenAI credentials
2. Add to repositories: Ensure the bot user has access to the repositories you want to monitor
3. Mention the bot: Use @your-bot-username in issue or MR comments to trigger responses
4. Get AI assistance: The bot will analyze the context and provide intelligent responses

Example Interactions

@gitbot Can you help me understand this bug report?
@gitbot What files are relevant to this feature request?
@gitbot Summarize the recent changes in this merge request

Client Certificate Authentication

GitBot supports client certificate authentication for OpenAI-compatible endpoints that require mTLS (mutual TLS) authentication:

export GITBOT_CLIENT_CERT_PATH=/path/to/client.crt
export GITBOT_CLIENT_KEY_PATH=/path/to/client.key
export GITBOT_CLIENT_KEY_PASSWORD=your_key_password  # Optional, for encrypted keys

Supported certificate formats:

• PKCS#12 (.p12, .pfx): Combined certificate and key file with password protection
• PEM: Separate certificate (.crt, .pem) and private key (.key) files

Note: The GITBOT_CLIENT_KEY_PASSWORD environment variable is only available as an environment variable for security reasons (no CLI argument).

🔍 Prompt Prefix Customization

GitBot allows you to add a consistent prefix to every prompt sent to the LLM. This is useful for:

• Setting a specific persona or role for the bot
• Adding formatting requirements
• Including security guidelines
• Specifying response length preferences

Usage

Set the prefix using either environment variable or command-line argument:

# Environment variable
GITBOT_PROMPT_PREFIX="You are an expert software developer. Always provide detailed explanations." ./gitbot

# Command-line argument
./gitbot --prompt-prefix "You are an expert software developer. Always provide detailed explanations."

Example

With this configuration:

GITBOT_PROMPT_PREFIX="You are a senior Python developer. Always include code examples in your responses."

When a user asks:

@gitbot How can I optimize this function?

The actual prompt sent to the LLM will be:

You are a senior Python developer. Always include code examples in your responses.

How can I optimize this function?

This ensures consistent behavior across all interactions.

Building from Source

# Clone and build
git clone https://github.com/myaple/gitbot.git
cd gitbot
cargo build

# Run tests
cargo test

# Run with development settings  
cargo run -- --help

Code Quality

This project maintains high code quality standards:

# Format code
cargo fmt

# Run linter  
cargo clippy

# Run all tests
cargo test

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines on:

• Code style and standards
• Testing requirements
• Pull request process
• Development workflow

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🆘 Support

• 📝 Create an issue for bug reports or feature requests
• 📖 Check the documentation for configuration help
• 💬 Join discussions in the Issues section

🏗️ Architecture

GitBot is built with:

• Language: Rust 🦀
• Async Runtime: Tokio
• HTTP Client: Reqwest
• AI Integration: OpenAI API
• Configuration: Clap + Environment Variables
• Logging: Tracing
• Testing: Built-in Rust testing + Mockito

---

Built with ❤️ by the GitBot team