📘 Advanced Multi-Agent Research System

An AI-powered research platform that simulates a structured academic workflow using RAG (Retrieval-Augmented Generation), FAISS vector memory, and a multi-agent architecture.

---

🚀 Overview

This system goes beyond traditional chatbots by:

• Planning before searching
• Retrieving real-time web data
• Generating structured research reports
• Critiquing and refining outputs
• Enabling chatbot interaction with generated reports

👉 It works like a “Chat with PDF” system, but dynamically generates the document first.

---

🧠 Key Features

• 🔍 Real-time web search using SerpAPI
• 🧩 RAG-based architecture
• 🗂️ Vector database using FAISS
• 🤖 Multi-agent system (Planner, Critic, Improver)
• 💬 Chatbot for report interaction
• ⚡ Parallel processing for faster execution
• 🎨 Streamlit-based UI

---

🏗️ System Architecture

Step-by-step pipeline:

1. User Input → Enter research topic
2. Planner Agent → Generates search queries
3. Web Retrieval → Fetches data using SerpAPI
4. Summarization → Extracts key insights
5. Embedding → Converts text into vectors
6. Vector Storage → Stored in FAISS
7. Context Retrieval → Top relevant chunks selected
8. Report Generation → Structured research report
9. Critic Agent → Identifies gaps
10. Improver Agent → Refines final output
11. Chatbot (RAG) → Answers questions from report

---

🧰 Tech Stack

• Python
• Streamlit
• OpenRouter API (LLM)
• SerpAPI (Web Search)
• FAISS (Vector Database)
• SentenceTransformers (Embeddings)
• NumPy

---

⚙️ Installation

1️⃣ Clone Repository

git clone <your-repo-link>
cd project-folder

2️⃣ Create Virtual Environment

python -m venv venv
venv\Scripts\activate

3️⃣ Install Dependencies

pip install streamlit faiss-cpu sentence-transformers openai python-dotenv requests numpy

4️⃣ Add Environment Variables

Create a .env file:

OPENROUTER_API_KEY=your_key_here
SERPAPI_API_KEY=your_key_here

---

▶️ Run the Application

streamlit run main.py

---

💡 How It Works

• The system retrieves real-time data from the web
• Converts it into embeddings using SentenceTransformers
• Stores embeddings in FAISS for similarity search
• Generates a research report using LLM
• Allows users to ask questions using a RAG-based chatbot

---

🎯 Advantages

• ✔ Reduces hallucination
• ✔ Uses real-time data
• ✔ Structured research output
• ✔ Interactive chatbot support
• ✔ Modular and scalable design

---

🔮 Future Scope

• 📚 Integration with academic sources (ArXiv, PubMed)
• 📄 PDF export (research paper format)
• 📑 Automatic citation generation (APA/IEEE)
• 🧠 Advanced RAG (re-ranking, hybrid search)
• ☁️ Cloud deployment

---

📜 License

This project is open-source and available under the MIT License.

---

❤️ Author

Developed as part of an AI research project.