RepoWise

RepoWise helps developers understand and contribute to unfamiliar GitHub repositories faster. Instead of spending hours exploring files and architecture manually, users receive automated documentation, onboarding guidance, code reviews, and contribution suggestions in minutes. The platform uses five collaborating AI agents powered by Band. Explorer analyzes repository structure and identifies important files. Documenter reads source code and generates technical documentation. Mentor converts that information into beginner-friendly onboarding guides. Reviewer evaluates documentation quality and provides scores, strengths, weaknesses, and recommendations. Task Suggester identifies beginner-friendly contribution opportunities. Unlike a simple pipeline, agents actively collaborate. If documentation quality scores below 50/100, Reviewer sends targeted feedback to Documenter for re-analysis and improvement. Agents communicate through Band's shared chat room using structured JSON messages over persistent WebSocket connections, with the full conversation history available for inspection. After analysis, users can chat directly with the Mentor agent and ask repository-specific questions about architecture, setup, workflows, and contribution paths. Answers are grounded in the generated repository knowledge. RepoWise is built with FastAPI, Next.js 14, Band SDK, Qwen via Featherless AI, and Groq as a fallback provider. Reliability features include session persistence, graceful degradation when external services fail, and a fallback mode that continues operating without API keys, making demos dependable under real-world conditions.