Trigent—RAG for Triaging GH Issues at Scale

Trigent enables Retrieval-Augmented Generation (RAG) over GitHub issues at scale. It fetches issues, enriches them with semantic embeddings, and provides an MCP server so AI agents can intelligently search, analyze, and triage large issue repositories.

What it does

• Fetches GitHub issues and pull requests from any repository
• Enriches them with semantic embeddings (using Mistral API)
• Provides an MCP server with tools for semantic search, similarity matching, and analytics
• Enables AI agents to perform intelligent issue triaging at scale

Quick Start

# Install and configure
pip install -e .
cp config.toml.example config.toml  # Add your Mistral API key

# Start Qdrant vector database
docker run -p 6333:6333 qdrant/qdrant
# or with Nix: services.qdrant.enable = true; (in configuration.nix)

# Setup a repository (fetches and enriches data)
trigent pull jupyterlab/jupyterlab

# Start MCP server for AI agent access  
trigent serve jupyterlab/jupyterlab

# Keep data updated
trigent update jupyterlab/jupyterlab

MCP Tools Available

The MCP server provides these tools for AI agents:

• get_issue(number) - Get specific issue details
• find_similar_issues(number) - Find semantically similar issues using embeddings
• find_similar_issues_by_text(text) - Find issues similar to given text
• find_cross_referenced_issues(number) - Get linked/referenced issues
• get_top_issues(sort_column) - Get top issues by any metric
• add_recommendation(issue_number, ...) - Add AI recommendations to issues

Other Commands

trigent export <repo>     # Export to CSV or visualizations  
trigent clean <repo>      # Remove repository data
trigent stats [<repo>]    # Show collection statistics

Configuration

Add your Mistral API key to config.toml:

[api]
mistral_api_key = "your_key_here"

Requirements

• Python 3.12+
• GitHub CLI (gh) for fetching issues
• Mistral API key for embeddings
• Qdrant vector database (runs locally by default)