Gatsby Agentic RAG Chatbot

A modular, agent-based Retrieval-Augmented Generation (RAG) system that answers questions about The Great Gatsby using the CrewAI framework, integrated with DeepEval for answer evaluation. This system demonstrates thoughtful architectural design, tool use, and autonomous reasoning across a multi-step pipeline.

---

Project Goals

• Design an agent-based RAG chatbot that answers literary questions based on a source document (e.g., The Great Gatsby).
• Use the CrewAI framework to model agents with tools, roles, goals, and backstories.
• Integrate an evaluator agent using DeepEval to score outputs based on:
  • Answer Relevance
  • Contextual Precision, Recall, and Relevancy
  • Faithfulness
• Store and retrieve context from a Qdrant vector store.
• Ensure modularity with reusable tools, agents, and pipeline logic.

---

Folder Structure

Agentic-RAG-Chatbot/
├── app/
│   ├── agents/              # Agent definitions (retriever, generator, evaluator)
│   ├── data/                # Source documents (e.g., The Great Gatsby)
│   ├── embeddings/          # Vector store utilities using Qdrant
│   ├── tools/               # Custom tools for retrieval and DeepEval
│   ├── pipeline.py          # Orchestrates the full multi-agent RAG pipeline
│   └── tasks.py             # Defines how tasks are passed between agents
│
├── main.py                  # Entry point to run the system interactively
├── requirements.txt         # Python dependencies
├── README.md                # (You are here)

---

High-Level Flow

[User Query]
     │
     ▼
[Retriever Agent]
     └── uses Qdrant vector DB to fetch top-k chunks
     │
     ▼
[Generator Agent]
     └── generates an answer grounded in retrieved context
     │
     ▼
[Evaluator Agent]
     └── uses DeepEvalTool to evaluate the output
     │
     ▼
[Final Output]
     └── includes: Answer and Evaluation Scores

---

Sample Output

---

Getting Started

Follow these steps to set up the Agentic-RAG-Chatbot locally.

Prerequisites

Make sure you have the following installed:

- Python 3.9+
- pip
- Docker (for Qdrant)
- Git

Clone the Repository

git clone https://github.com/SameerSRao/Agentic-RAG-Chatbot.git
cd Agentic-RAG-Chatbot

Set Up Environment

Using virtual environment:

python3 -m venv venv
source venv/bin/activate   # On Windows: venv\Scripts\activate
pip install -r requirements.txt

Configure Environment Variables

cp .env.example .env

Start Qdrant

docker pull qdrant/qdrant  #install qdrant if not installed locally
docker run -p 6333:6333 -p 6334:6334 qdrant/qdrant

Run App

python main.py