Pokémon Agentic RAG Assistant

An intelligent AI assistant built to analyze and retrieve Pokémon data using a sophisticated Agentic RAG (Retrieval-Augmented Generation) architecture. This system intelligently routes queries between semantic search and structured data analysis.

🧠 Key Architecture: The Agentic Router

This project implements an Agentic Router that classifies user intent to provide the most accurate response:

• FAISS Route: Handles semantic, descriptive, and background queries by retrieving context from a vector database.
• Pandas Route: Handles statistical, mathematical, and comparative queries by executing dynamic analysis on the structured dataset.

🛠 Tech Stack

• Backend: Flask (Python)
• AI/ML: Google Gemini (LLM), FAISS (Vector DB), Sentence-Transformers
• Data Processing: Pandas, NumPy
• Infrastructure: Docker, AWS EC2

🚀 Deployment

The application is containerized and ready for production deployment.

docker run -p 5001:5001 \
  -e GEMINI_API_KEY="your_api_key_here" \
  --name pokemon-rag \
  adammes/pokemon-rag:1.0

📂 Project Structure

• app.py: Main application logic, including the Agentic Router and RAG chains.
• data/: Contains the source "Pokemon.csv" file used as the knowledge base.
• static/ & templates/: Web interface assets and HTML layouts.
• Dockerfile: Configuration for building the container image (linux/amd64).
• requirements.txt: List of Python dependencies.

📊 Data Credit

The dataset used in this project was sourced from Kaggle.

📸 Screenshots

Faiss route

Pandas route