RAG Document Search

An AI-powered document search application that allows users to upload files and ask natural language questions about their content.

The system uses Retrieval-Augmented Generation (RAG) with semantic search to find relevant information and generate accurate answers.

Features

• Upload PDF, DOCX, and TXT files
• Automatic text extraction and chunking
• Semantic search using vector embeddings (pgvector)
• AI-generated answers based on document content
• Source referencing for retrieved information
• Document management (view, download, delete)
• PDF preview and full-text viewer

Tech Stack

• Frontend: Next.js, React, Tailwind CSS
• Backend: Next.js API Routes
• Database: Supabase (PostgreSQL + pgvector)
• AI: OpenAI (embeddings + LLM)
• Storage: Supabase Storage

How It Works

1. Upload

   • Files are uploaded and stored in Supabase Storage
   • Text is extracted and split into chunks

2. Embedding

   • Each chunk is converted into a vector using OpenAI embeddings

3. Storage

   • Chunks + embeddings are stored in PostgreSQL (pgvector)

4. Search

   • User query is embedded
   • Similar chunks are retrieved using vector similarity

5. Generation

   • Retrieved context is passed to an LLM
   • Final answer is generated

Setup

1. Clone the repo

git clone <your-repo-url>
cd <your-project>

2. Install dependencies

npm install

3. Add environment variables

Create a .env.local file in the root directory:

NEXT_PUBLIC_SUPABASE_URL=your_supabase_url
NEXT_PUBLIC_SUPABASE_PUBLISHABLE_DEFAULT_KEY=your_supabase_anon_key
SUPABASE_SERVICE_ROLE_KEY=your_service_role_key
OPENAI_API_KEY=your_openai_api_key

4. Run the app locally

npm run dev