Chat with your docs. Get cited answers.
Upload any PDF. Ask questions in plain English. AdvChat finds the exact source — every time.
How it works
Three steps to insights
Upload PDF
Drag and drop any PDF up to 25MB. AdvChat chunks and indexes it instantly.
Ask in plain English
Type your question naturally. No special syntax or commands needed.
Get cited answers
Receive accurate answers with exact page references you can verify.
Capabilities
Built for real use cases
Source Citations
Every answer links to the exact page in your document.
Real-time Streaming
Answers stream token by token via Server-Sent Events.
Multi-document
Upload multiple PDFs and query across all of them.
Redis Caching
Repeated questions return instantly from cache.
Secure Upload
PDF-only, 25MB limit, files never publicly exposed.
Configurable
Chunk size, overlap, top-K — all tunable via env vars.
Under the hood
Production-grade RAG pipeline
<2s
First token latency
Redis caching on repeated queries returns answers instantly.
Pinecone
1536-dim vector search, namespace-scoped per user.
SSE Streaming
Real-time token streaming via Server-Sent Events from Express.
Production-ready infrastructure
Node.js cluster, graceful shutdown, Zod validation, Winston structured logging.