CodeStory

'Why does this code exist?' is the #1 question developers ask when reading unfamiliar code. Developers spend about a third of their time on code maintenance, not new features. git blame shows who and when, but not why. Legacy code is hard to understand, onboarding takes weeks of tribal knowledge, and decisions are buried in 100+ commits and closed issues.

AI-powered git archaeology that tells the story behind your code. The system analyzes blame, commit history, and GitHub issues, then uses Llama 3.2 to turn raw history into readable narratives. An interactive timeline shows origin → refactors → bug fixes → today, highlighting patterns like hotfixes, major refactors, and key contributors.

• Hit a confusing function
• Run CodeStory with the repo, file, and function
• In about 30 seconds, get a narrative like: 'Retry logic added after prod outage #342 (Dec 2023), refactored by Mike (Jan 2024) for better error handling'
• View the timeline plus linked issues and PRs
• Instantly know what happened and who to ask

• History Tracer agent — walks through commits
• Context Gatherer agent — pulls related issues and PRs
• Story Generator agent — produces narrative + timeline
• LLM: Llama 3.2 (via Ollama or Groq)

• Backend: Python + FastAPI
• Git: PyGit2 (fast history + blame)
• LLM runtime: Ollama (Llama 3.2)
• Context: GitHub REST API
• Frontend: React
• Optional: Docker
• Vibe coding: Cursor, VS Code + Copilot, Claude, ChatGPT, Gemini