docai

Project Overview

docai is an AI-powered document Q&A and retrieval-augmented generation (RAG) system. It allows users to upload documents, process them, and interact with them using state-of-the-art LLMs (Gemini, GPT-4, Claude, etc.) via a modern web interface.

Demo

Features

• Connect and process documents using Google docs
• Semantic search and context retrieval (RAG)
• Chat interface with LLM responses
• Model selection (Gemini, GPT-4, Claude, ...)
• File system-based vector/document storage

---

Local Setup Instructions

1. Prerequisites

• Node.js (v18+ recommended)
• npm or yarn
• A Google AI API key (for Gemini)
• (Optional) API keys for Cohere, OpenAI, or Anthropic if using other models

2. Clone the Repository

git clone https://github.com/apsinghdev/docai.git
cd docai

3. Install Dependencies

pnpm install

4. Environment Variables

Create a .env.local file by running cp .env.example .env.local in the root directory and add the mentioned credentials.

5. Run the Development Server

pnpm dev

The app will be available at http://localhost:3000

---

- `lib/` — Core logic for LLMs, embeddings, RAG, and storage
- `app/api/` — Next.js API routes for chat and document processing
- `components/` — React components for the frontend UI

---

Architecture Diagram

User Interface (React/Next.js)
           |
           v
  API Routes (Next.js)
           |
           v
   LangChain RAG System
   /        |         \
  v         v          v
File   Embeddings     LLMs
System   (Cohere)   (GPT-4, etc.)
Storage                |
   \        |         /
           v
   LangChain RAG System
           |
           v
  API Routes (Next.js)
           |
           v
User Interface (React/Next.js)

• User Interface: Users upload documents and chat with the AI.
• API Routes: Handle requests, route to RAG logic or document processing.
• LangChain RAG System: Retrieves relevant context, manages embeddings, and orchestrates LLM calls.
• LLMs: Gemini (Google), GPT-4 (OpenAI), Claude (Anthropic), etc.
• Embeddings: Used for semantic search and context retrieval.
• File System Storage: Stores processed document chunks and embeddings.

---