🚀 Production-ready multi-agent AI system that decomposes complex queries into structured workflows using LangGraph, Groq LPU, and Tavily Search API
- 🤖 Autonomous Multi-Agent System with task decomposition
- ⚡ 500+ tokens/sec inference using Groq LPU
- 🔍 Real-time web search via Tavily API
- 🎯 Reduced hallucinations with live retrieval
- 💻 Streamlit UI with real-time execution tracking
- 🆓 Optimized for free-tier usage (12K TPM)
- Research Agent 🔍 → Retrieves data via Tavily
- Critique Agent 🔎 → Validates and improves results
- Summarizer Agent 📝 → Generates final structured output
User Query
↓
Research Agent (Search + LLM)
↓
Critique Agent (Validation Loop)
↓
Decision Layer
├── More Research → Loop Back
└── Finalize → Summarizer
↓
Final Response
| Layer | Technology |
|---|---|
| Orchestration | LangGraph |
| LLM Inference | Groq (Llama-3.3-70B) |
| Retrieval | Tavily Search API |
| Backend | Python |
| UI | Streamlit |
git clone https://github.com/rjkalash/multi-agent-research-assistant.git
cd multi-agent-research-assistantpython -m venv venv
# Windows
venv\Scripts\activate
# macOS/Linux
source venv/bin/activatepip install -r requirements.txtCreate .env file:
GROQ_API_KEY=your_key_here
TAVILY_API_KEY=your_key_herepython main.py "Latest AI breakthroughs in 2026"streamlit run app.py| Metric | Value |
|---|---|
| Speed | 500+ tokens/sec |
| Latency | 10–30 sec |
| Token Usage | ~7K/query |
| Reliability | High |
| Hallucination | Reduced (via retrieval) |
- Latest AI breakthroughs in 2026
- Quantum computing advancements
- Climate change impact on agriculture
- Cybersecurity trends
multi-agent-research-assistant/
├── agents.py
├── app.py
├── main.py
├── utils.py
├── demo.py
├── requirements.txt
├── .env.example
└── docs/
PRs are welcome!
Raj Kalash Tiwari
- GitHub: https://github.com/rjkalash
✅ Production-ready system for real-world AI workflows
⭐ Star this repo if you found it useful!