DocoChat AI — RAG-Powered Document Conversation Platform

A document chat app built with Retrieval-Augmented Generation (RAG). Upload documents, train them with AI, and chat to extract insights.

What It Does

Upload documents (PDF, DOCX, TXT, MD, RTF, CSV), train them using semantic embeddings, and ask questions in natural language. The AI answers using your document content, not generic knowledge.

How It Works (Technical Overview)

RAG Pipeline:

1. Document processing: extracts text from multiple formats

2. Intelligent chunking: splits text into 1000-character chunks with 20-character overlap

3. Vector embeddings: converts chunks into 1536-dimensional vectors using Google Gemini's embedding-001

4. Vector storage: stores embeddings in Supabase PostgreSQL with pgvector

5. Semantic search: uses cosine similarity to find relevant document sections

6. Contextual responses: Gemini 2.0-flash-lite generates answers from retrieved context

Architecture:

• Frontend: Next.js 15 (App Router) + React 19 + TypeScript

• UI: TailwindCSS 4 + shadcn/ui

• Backend: Next.js API Routes + Server Actions

• Database: Supabase (PostgreSQL + pgvector)

• AI: Google Gemini (embeddings + chat)

• Processing: LangChain text splitters

Key Features

• Multi-format support: PDF, DOCX, TXT, MD, RTF, CSV

• Semantic search: meaning-based retrieval, not keyword matching

• Conversation memory: maintains chat history

• Credit system: built-in usage tracking (10 free credits)

• Modern UI: glassmorphism design, dark/light mode, responsive

• Privacy-first: documents stored in your Supabase instance

Technical Highlights

• Type-safe: TypeScript with strict checking

• Error handling: validation and error boundaries

• Performance: optimized vector search with indexed queries

• Scalable: stateless architecture, horizontal scaling ready

• Production-ready: error handling, validation, security practices

Use Cases

• Research: query academic papers and research documents

• Documentation: interact with technical docs and manuals

• Business: analyze reports, contracts, and business documents

• Education: study materials and lecture notes

• Personal: organize and search personal documents

Built with modern web technologies and production-ready architecture. Try it with 10 free credits.

Live Demo: docochat-rag-ai.hemantsharma.tech