Project Mind — Turn Your GitHub History Into Searchable Memory

This Hacktoberfest Weekend Challenge submission, called Project Mind, transforms GitHub repositories into searchable memory systems powered by AI. Rather than searching through scattered documentation, issues, pull requests, and commit histories, users can simply ask Project Mind questions in natural language. The AI retrieves relevant parts of the actual project code and history and explains the flow with source references.

Project Mind indexes various elements from GitHub repositories including source code, README and Markdown docs, GitHub issues, pull requests, commit history, approved long-term memories, and project decisions and notes. When a question is posed, it combines vector search and keyword search techniques to locate the most relevant contextual information. The retrieved data is then passed to a locally running Llama 3.2 3B model for natural language generation of an answer grounded in the project's own codebase and history.

One key differentiator is that Project Mind focuses on understanding "why" a project works a certain way, rather than just providing code examples on how to implement features. It aims to fill the gap between generic AI coding tools and more specialized developer assistance by preserving not just "what" the code does, but also "why" decisions were made during development.

Demo videos showcase the full workflow: connecting a GitHub repo, indexing its knowledge, browsing indexed data, asking questions like how GitHub authentication works, tracing back relevant code and commits, searching project history for past bugs and decisions, and even saving important project knowledge for future queries. The codebase leverages open-weight AI models running locally through Ollama for privacy and security reasons, avoiding sending proprietary project data to third-party services.