laurentongaro/odinrag

OdinRAG is a personal knowledge base for learning and working with the Odin programming language, specifically for game development topics like Raylib, Sokol, hot-reload patterns, and memory allocators. Rather than a runnable program, it is a collection of scrapers, agent skills, and document indexes that help you gather, format, and query learning material about Odin game development. The project supports two styles of navigation: a human browsing through an Obsidian vault or reading INDEX.md files, and an AI coding agent using a set of Kilo Code skills and a custom subagent definition to answer questions about Odin game development. The AI layer uses a RAG approach, meaning the agent can look up specific documents in the knowledge base before answering rather than relying only on its trained knowledge. The repo ships only the workflow infrastructure, not the actual content. Copyrighted material such as a paid Odin book and Skool course transcripts is intentionally excluded. Python scrapers are included for collecting publicly available documentation from official sources, and a separate scraper handles paid content if you have your own subscription. A formatting script cleans up Odin code blocks found throughout the Markdown files. Configuration lives in a single gitignored file so personal paths and credentials never end up in version control. The project uses a split-branch approach: the public GitHub branch contains only sanitized content, while the author maintains a private branch with personal notes and paywalled material that is never pushed to GitHub. The entire project structure, including the scraper architecture, agent prompts, and skill definitions, was designed collaboratively with the MiniMax-M3 AI model through the Kilo Code IDE. The code is released under MIT, the scraped content belongs to its original authors.