📄 Multi-Utility RAG AI Agent using LangGraph & Streamlit

A production-style Retrieval Augmented Generation (RAG) chatbot built using LangGraph, LangChain, Cerebras (Qwen), FAISS, and Streamlit.

This chatbot can:

• 💬 Chat normally with memory
• 📄 Answer questions from uploaded PDFs (RAG)
• 🔧 Use tools like Web Search, Calculator, and Stock Price Fetching
• 🧵 Maintain multiple chat threads
• 💾 Persist conversations using SQLite

---

🚀 Features

🧠 Agentic AI (LangGraph)

• Graph-based agent workflow
• Tool calling with conditional routing
• Stateful conversations with persistence

📚 PDF-based RAG

• Upload PDFs per chat thread
• Chunking + embeddings using HuggingFace
• FAISS vector search for fast retrieval
• Thread-wise document indexing

🔧 Built-in Tools

• 🔍 Web Search (DuckDuckGo)
• 🧮 Calculator (add, sub, mul, div)
• 📈 Stock price lookup (Alpha Vantage API)
• 📄 RAG tool for PDF-based Q&A

💬 Chat Experience

• Streaming responses (token-by-token)
• Multi-threaded conversations
• Resume old chats
• Thread-specific memory & documents

💾 Persistence

• SQLite-based checkpointing
• Conversations survive app restarts
• Thread history is stored automatically

---

🏗️ Tech Stack

Layer	Technology
LLM	Cerebras Qwen-3-32B
Agent Framework	LangGraph
LLM Interface	LangChain
Embeddings	HuggingFace (MiniLM)
Vector DB	FAISS
UI	Streamlit
Persistence	SQLite
Tools	DuckDuckGo, Alpha Vantage

---

📂 Project Structure

rag-ai-agent/
│
├── rag_agent_backend.py     # LangGraph agent, tools, RAG logic
├── rag_agent_frontend                   # Streamlit frontend
├── chatbot.db               # SQLite database (auto-created)
├── .env                     # API keys & environment variables
├── requirements.txt         # Python dependencies
└── README.md                # Project documentation

---

🔄 High-Level Workflow

User Message
   ↓
LangGraph Chat Node
   ↓
(If needed)
Tool Selection
   ├── Web Search
   ├── Calculator
   ├── Stock Price
   └── RAG Tool (PDF)
   ↓
Tool Result
   ↓
LLM Response
   ↓
Persist State (SQLite)

---

📄 How PDF RAG Works

1. User uploads a PDF

2. PDF is:

   • Loaded using PyPDFLoader
   • Split into chunks
   • Embedded using HuggingFace embeddings

3. Chunks are stored in FAISS

4. Each chat thread has its own retriever

When a question is asked:

• Relevant chunks are retrieved
• Context is injected into the LLM
• Answer is generated

---

🧠 Thread-Based Memory

• Each chat has a unique thread ID

• Threads store:

  • Conversation history
  • Indexed PDFs

Users can:

• Start a new chat
• Resume old chats
• Use different PDFs in different threads

---

🛠️ Tools Available

🧮 Calculator Tool

• Operations: add, sub, mul, div

🔍 Web Search Tool

• Powered by DuckDuckGo

📈 Stock Price Tool

• Uses Alpha Vantage API

📄 RAG Tool

• Retrieves relevant content from uploaded PDFs

---

▶️ How to Run the Project

1️⃣ Create virtual environment

python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

2️⃣ Install dependencies

pip install -r requirements.txt

3️⃣ Set environment variables

Create a .env file:

CEREBRAS_API_KEY=your_api_key_here

4️⃣ Run the Streamlit app

streamlit run app.py

---

📌 Example Use Cases

• Chat with your college notes PDFs
• Ask questions from research papers
• Financial queries using stock price tool
• Mixed conversations using tools + RAG
• Long-term chatbot memory

---

🧩 Future Enhancements

• 🔐 Authentication
• 📊 LangSmith observability
• 🧠 Hybrid RAG (BM25 + Vector)
• 🗂️ Multi-document comparison
• 🧪 Evaluation & feedback loop
• 🌐 Deployment (Docker / Cloud)

---

🙌 Acknowledgements

• LangChain & LangGraph team
• Cerebras Systems
• HuggingFace
• Streamlit community

---

⭐ If you like this project

Give it a ⭐ on GitHub and feel free to fork & extend it!