🤖 Agentic RAG Chatbot

A Retrieval-Augmented Generation (RAG) chatbot powered by an agentic architecture using LlamaIndex, OpenAI GPT-4o, and Streamlit. This system enables dynamic interaction with multiple PDFs (like earnings reports or research papers), where an LLM intelligently decides whether to query documents or generate answers directly.

✨ Key Features

• Agent-based Orchestration: Uses FunctionCallingAgentWorker and AgentRunner to dynamically manage document tools and query resolution.
• Multi-Document Handling: Upload and chat with multiple PDFs through a modular Streamlit interface.
• Tool Indexing with LlamaIndex: Each document is transformed into vector-based and summarization tools using VectorStoreIndex and SummaryIndex.
• Smart Query Routing: The agent retrieves the right tool (vector or summary) depending on the user query context.
• Document Caching & Frontend: Interactive document upload and chat interface using Streamlit with persistent document storage.

📂 Project Structure

agentic-rag-chatbot/
├── Scripts/
│   ├── app.py                  # Main backend logic (agent construction, doc processing)
│   ├── frontend.py             # Streamlit-based UI
│   ├── utils.py                # Config reader
│   └── cache/                  # Uploaded documents (PDFs)
├── LICENSE
├── .gitignore

🚀 How to Run

1. Clone & Setup Environment

git clone https://github.com/MThabsheer7/agentic-rag-chatbot.git
cd agentic-rag-chatbot
python -m venv venv
source venv/bin/activate   # or .\venv\Scripts\activate on Windows
pip install -r requirements.txt

Note: Set your OpenAI API key in the YAML file referenced in utils.py.

2. Run the App

cd Scripts
streamlit run frontend.py

Upload multiple PDF documents through the sidebar, assign them names, and then interact via the chat interface.

📦 Dependencies

• llama-index (LlamaIndex)
• streamlit
• openai
• PyYAML
• loguru

Install via:

pip install -r requirements.txt

📁 Config Example (config.yaml)

api:
  openai: sk-xxxxxxxxxxxxxxxxxxxx

🧠 Use Case

This system was built to support document-intensive domains like financial reporting, where users can upload multiple earnings reports or company filings and ask complex, cross-document questions using a single conversational agent.