📘 Advanced Multi-Agent Research System

An AI-powered research platform that simulates a structured academic workflow using Retrieval-Augmented Generation (RAG), FAISS Vector Memory, Multi-Agent Architecture, and T5-based Summarization.

---

🚀 Overview

This system goes beyond traditional chatbots by:

• Planning before searching
• Retrieving real-time web information
• Generating professional research reports
• Critiquing and refining outputs
• Verifying factual accuracy
• Creating executive summaries
• Exporting reports as PDFs
• Enabling interactive Q&A on generated reports

👉 Think of it as a dynamic "Chat with PDF" system that first creates the research document and then allows users to interact with it intelligently.

---

🧠 Key Features

• 🔍 Real-time web search using SerpAPI

• 🧩 Retrieval-Augmented Generation (RAG)

• 🗂️ FAISS Vector Database for semantic retrieval

• 🤖 Multi-Agent System

  • Planner Agent
  • Writer Agent
  • Critic Agent
  • Improver Agent
  • Verifier Agent
  • Summarizer Agent

• 📋 T5-based Executive Summary Generation

• 📥 Professional PDF Export

• 🎓 Quality Assessment & Confidence Scoring

• 💬 Interactive RAG-Powered Q&A Assistant

• ⚡ Parallel Search Processing

• 🎨 Streamlit-Based User Interface

• 📚 Semantic Memory using Sentence Transformers

---

🏗️ System Architecture

---

🤖 Agent Responsibilities

📌 Planner Agent

• Analyzes the research topic
• Generates optimized search queries
• Identifies important focus areas

🔍 Retrieval Layer

• Uses SerpAPI for real-time information retrieval
• Collects relevant data from multiple sources

💾 Memory Layer

• Splits retrieved content into chunks
• Creates embeddings using Sentence Transformers
• Stores vectors in FAISS for semantic search

📝 Writer Agent

• Generates structured research drafts
• Organizes content into logical sections

🔎 Critic Agent

• Reviews generated reports
• Detects missing information
• Suggests improvements

✨ Improver Agent

• Refines content quality
• Enhances clarity and completeness

✅ Verifier Agent

• Performs fact-checking
• Validates critical information
• Improves report reliability

📋 Summarizer Agent (T5)

• Generates concise executive summaries
• Extracts key insights
• Produces reader-friendly content

💬 Q&A Assistant

• Retrieves relevant context from FAISS
• Answers user questions using RAG
• Enables interactive report exploration

---

🧰 Tech Stack

Programming & Frameworks

• Python
• Streamlit

AI & Machine Learning

• Llama 3 8B Instruct (OpenRouter)
• T5 Transformer
• Sentence Transformers
• Retrieval-Augmented Generation (RAG)

Search & Retrieval

• SerpAPI
• FAISS Vector Database

Libraries

• NumPy
• Requests
• Python Dotenv

---

⚙️ Installation

1️⃣ Clone Repository

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

2️⃣ Create Virtual Environment

python -m venv venv

Activate Environment:

Windows

venv\Scripts\activate

Linux / Mac

source venv/bin/activate

3️⃣ Install Dependencies

pip install -r requirements.txt

4️⃣ Configure Environment Variables

Create a .env file:

OPENROUTER_API_KEY=your_api_key
SERPAPI_API_KEY=your_api_key

---

▶️ Run the Application

streamlit run main.py

---

🎯 Advantages

✔ Reduces hallucinations using RAG

✔ Uses real-time web information

✔ Semantic retrieval with FAISS

✔ Multi-agent quality improvement workflow

✔ Fact-checking through Verifier Agent

✔ Executive summaries with T5

✔ Professional PDF export

✔ Interactive report chatbot

✔ Modular and scalable architecture

---

🔮 Future Scope

• 📚 Integration with ArXiv and PubMed
• 📑 Automatic Citation Generation (APA / IEEE)
• 🔄 Hybrid Search (BM25 + Vector Search)
• 🎯 Re-ranking Models
• 🌐 Multi-Language Research Support
• ☁️ Cloud Deployment (AWS, Azure, GCP)
• 📊 Research Analytics Dashboard
• 🧠 Advanced Multi-Agent Collaboration

---

❤️ Author

Ankita Ghavate