DeepSyncAI is an intelligent multi-agent Web RAG research engine that autonomously searches the web, extracts relevant information from live sources, generates comprehensive LLM-powered research reports, and performs AI-driven critique, all from a single topic input.
Designed with a modern agentic RAG architecture, DeepSyncAI orchestrates a sequential four-stage pipeline consisting of a real-time search agent, a deep extraction agent, an AI report generation agent, and an intelligent LLM critic agent. The platform delivers structured Markdown outputs through a sleek four-tab interface with smooth animations, typewriter-style rendering, session history tracking, and a production-grade glassmorphism UI, creating a fast, interactive, and research-focused AI experience.
DeepSyncAI Live: https://deepsyncai.vercel.app
- Multi-Agent Research Pipeline (Search → Extract → Report → Critic)
- RAG (Retrieval-Augmented Generation) Architecture
- LLM-Powered Report Generation & Critique (Gemini)
- Tavily-Powered Real-Time Web Search
- Deep Structured Content Extraction from Live Web Sources
- Gemini AI Research Report Generation
- AI-Powered Critic with Accuracy Scoring & Insightful Feedback
- Four-Tab Result Interface (Search / Extract / Report / Critic)
- Auto-Switching Tabs as Each Agent Completes
- Interactive Research History with Full Session Replay
- Structured & Formatted Markdown Output Rendering
- Rich Code Blocks with Copy Support
- Fast Typewriter-Style Output Animation
- Smooth UI Animations (Framer Motion)
- Fully Responsive Design
- Modern Dark UI with Black & Blue Glassmorphism Theme
- Real-Time Toast Notifications per Agent Step
- Client-Side Safe API Handling
| Technology | Category | Purpose |
|---|---|---|
| Next.js (App Router) | Frontend Framework | Full-stack React framework for routing, SSR, and API handling |
| LLM (Large Language Model) | AI Concept | Gemini used as the core reasoning engine for report and critique generation |
| RAG (Retrieval-Augmented Generation) | AI Architecture | Retrieved web data injected as context into Gemini before generation |
| React 19 | UI Library | Component-based user interface development |
| Next.js API Routes | Backend APIs | Secure server-side agent orchestration and AI request handling |
| Tailwind CSS 4 | Styling and Responsiveness | Utility-first styling with a modern, responsive design system |
| Framer Motion | Animations | Smooth, performant UI animations, tab transitions, and micro-effects |
| React Markdown | Markdown Rendering | Structured rendering of AI-generated research output |
| Remark GFM | Markdown Extensions | GitHub-Flavored Markdown support (tables, lists, code blocks) |
| React Hot Toast | Notifications | Real-time per-step toast notifications and user feedback |
| Google Generative AI (Gemini) | AI Integration | Report generation, content synthesis, and intelligent critique |
| Tavily API | Web Search | Real-time web search with URL and content retrieval |
| Cheerio | Web Scraping | Server-side HTML parsing and content extraction from live URLs |
| TypeScript | Type Safety | Static typing and a safer, maintainable codebase |
| React Icons | Icons | Consistent, scalable icon system |
| Vercel | Deployment & Hosting | Fast, global deployment with optimized performance |
User Input (Topic)
│
▼
┌─────────────┐
│ Search │ Tavily API: fetches URLs + content snippets [Retrieval]
└──────┬──────┘
│
▼
┌─────────────┐
│ Extract │ Cheerio: scrapes full content from top URLs [Retrieval]
└──────┬──────┘
│
▼
┌─────────────┐
│ Report │ Gemini LLM: synthesizes report grounded on retrieved context [RAG + LLM]
└──────┬──────┘
│
▼
┌─────────────┐
│ Critic │ Gemini LLM: scores report, lists strengths & weaknesses [RAG + LLM]
└─────────────┘
- Clone the repository
git clone https://github.com/OmkarArdekar12/DeepSyncAI.git
cd DeepSyncAI- Install dependencies
npm install- Set up environment variables
- Create a
.env.localfile in the root directory and add your API keys
GEMINI_API_KEY=your_gemini_api_key
TAVILY_API_KEY=your_tavily_api_key
- Run the development server
npm run dev