SlashDocs

• 

  Documentation made easier.

Inspiration 🧠

When developers are initially hired and onboarded by their companies, they are often assigned and expected to navigate massive codebases on their own. This process, without any assisting software can take months.

Large companies like Google have taken to building internal tools that index their codebases, automatically generate documentation, and provide contextual chat with LLMs using RAG. On the other hand, there's software like DeepWiki by Devin that provides this service to the public.

But what happens when a leaner team can't afford to build their own self-maintaining documentation, but also isn't willing to hand over proprietary code?

Similar to how Supabase offers its services as opposed to Firebase, we aim to solve this problem by offering an open-source, self-hostable documentation platform for developers and enterprise companies alike.

What it does

• Clone and index any GitHub repository
• Automatically preprocess and embed key files into ChromaDB
• Query codebases in natural language using a Retrieval-Augmented Generation (RAG) pipeline
• Navigate sections through a clean, wiki-inspired Next.js frontend with smooth scrolling and instant linking

How we built it 🏗️

               Tech Stack:
• **Frontend**: Next.js, TailwindCSS
• **Backend**: FastAPI (Python)
• **Database**: ChromaDB (vector embeddings)
• **AI / LLM**: OpenAI text-embedding-3-small
• **Tools**: GitHub, Postman, Pytest, dotenv

Challenges we ran into ⚔️

• Chunk Sizing: Finding the right balance between embedding chunk size and retrieval accuracy was tricky — small chunks improved precision but increased token usage, while large chunks hurt context quality.
• Building the RAG Pipeline: Integrating preprocessing, embedding, storage, and query retrieval into a seamless pipeline took significant debugging and iteration.
• CORS & Environment Variables: Handling cross-origin requests and securely loading .env variables across FastAPI and Next.js.
• Frontend Navigation: Implementing smooth, in-panel scrolling and dynamic linking between documentation sections.

Accomplishments that we're proud of 🏅

• Built a fully functional RAG pipeline within 15 hours
• Achieved seamless interaction between Next.js and FastAPI
• Designed a responsive, minimalist UI inspired by DeepWiki
• Indexed and queried multiple public GitHub repositories successfully
• Learned how to combine LLMs + vector databases for practical use cases

What we learned 📚

• Deep understanding of RAG concepts
• How ChromaDB handles embeddings and similarity search
• How to structure modular APIs for scalable RAG pipelines
• The importance of good UI/UX in knowledge discovery tools

What's next for SlashDocs 🚀

• Add team-based repository sharing and versioning
• Support multi-repo embeddings and comparison queries
• Add real-time chat mode for conversational repo exploration
• Deploy an open-source community platform for documentation search

Built With

• chromadb
• fastapi
• next.js
• openai
• tailwindcss