RTFM

An AI powered documentation assistant built with RAG (Retrieval Augmented Generation). Upload your documentation files, ask questions in natural language, and get accurate answers with source citations, powered entirely by free-tier APIs.

Tech Stack

Layer	Technology
Backend	NestJS (TypeScript)
Frontend	React + Vite + TailwindCSS
Vector DB / Cache / Storage	Redis Stack
LLM	Groq (llama-3.3-70b-versatile) with Google AI (Gemini) fallback
Embeddings	Jina AI (jina-embeddings-v3, 1024 dimensions)
Package Manager	pnpm

How It Works

Ingestion

1. upload .md or .txt documentation files
2. split content into chunks by markdown headers and word count
3. generate 1024-dim embeddings via Jina AI
4. store chunks + embeddings in Redis vector index (idx:docs)
5. deduplicate via SHA-256 content hash

Question Answering (RAG)

1. embed the user question via Jina AI (retrieval.query task)
2. check semantic cache (idx:cache) for a similar previously answered question
3. if cache miss, run KNN vector search to find top 5 relevant chunks
4. build prompt with retrieved context and send to Groq LLM
5. return answer with source file citations
6. cache Q&A pair for future similar questions

Quick Start

1. Get API Keys (All Free)

• Groq : LLM (free tier: 14k requests/day)
• Jina AI : Embeddings (free tier: 1M tokens)
• Google AI Studio : Fallback LLM (free tier: 1M tokens/day)

2. Setup Environment

cd server
cp .env.example .env
# Fill in your API keys in .env

3. Start Redis

docker-compose -f docker/docker-compose.yml up -d

4. Start Backend

cd server
pnpm install
pnpm run start:dev

5. Start Frontend

cd client
pnpm install
pnpm run dev

App runs at http://localhost:5173, backend at http://localhost:3000.

API Endpoints

Endpoint	Method	Description
/api/ingest	POST	Upload documentation files (multipart)
/api/ask	POST	Ask a question { question: string }
/api/documents	GET	List all uploaded documents
/api/documents/:id	DELETE	Delete a document and its chunks

Redis Data Model

Key Pattern	Type	Description
file:<hash>	Hash	Document metadata (fileName, chunks, createdAt)
doc:<hash>:<i>	Hash	Document chunk with embedding vector
cache:<id>	Hash	Cached Q&A pair with embedding vector
idx:docs	Index	Vector index for document chunk search
idx:cache	Index	Vector index for semantic cache lookup

UI preview (V1)

License

MIT