FriendsRAG 📺

A Retrieval-Augmented Generation (RAG) system that helps Friends fans find episodes based on scene descriptions. Simply describe any scene from the show, and FriendsRAG will identify the episode and provide a complete summary!

Example:

• You ask: "What's the episode where Rachel gets off the plane?"
• FriendsRAG returns: Episode title, season, episode number, and full summary

📁 Repository Structure

FriendsRAG/
├── RAG.ipynb                # Core RAG implementation
├── RAG_Langfuse.ipynb       # RAG with Langfuse monitoring
├── RAG_FastAPI.py           # FastAPI endpoint
├── requirements.txt         # Python dependencies
└── data/
    ├── FriendsEpisodes.csv  # Episode metadata
    └── FriendsScripts.csv   # Episode scripts

🚀 Quick Start

1. Install Dependencies

Using uv (recommended):

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

Using pip:

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

2. Configure AWS Credentials

Login with AWS SSO:

aws sso login --profile your-profile-name

Update the AWS configuration in each file:

REGION_NAME = "us-east-1"  # Your AWS region
CREDENTIALS_PROFILE_NAME = "your-profile-name"  # Your AWS profile

📖 Usage Guide

RAG.ipynb - Core Implementation

What it does: Creates the vector database and allows you to query episodes interactively.

Requirements:

• AWS credentials configured
• Data files in data/ folder

How to run:

jupyter notebook RAG.ipynb

Run all cells to:

1. Load episode and script data
2. Create embeddings and build FAISS vector database
3. Query the system with natural language

---

RAG_Langfuse.ipynb - With Monitoring

What it does: Same as RAG.ipynb but adds Langfuse monitoring and evaluation metrics.

Requirements:

• AWS credentials configured
• Data files in data/ folder
• Langfuse account and API keys

Setup Langfuse:

# Create a .env file with:
LANGFUSE_PUBLIC_KEY=your_public_key
LANGFUSE_SECRET_KEY=your_secret_key
LANGFUSE_BASE_URL = "https://cloud.langfuse.com"

How to run:

jupyter notebook RAG_Langfuse.ipynb

Features:

• Request/response tracing
• Evaluation metrics (correctness, relevance, groundedness)
• Dashboard at https://cloud.langfuse.com

---

RAG_FastAPI.py - API Endpoint

What it does: Provides a REST API to query the RAG system.

Requirements:

• AWS credentials configured
• Vector database already created (run RAG.ipynb first to generate ./vector_database/)

How to run:

fastapi dev RAG_FastAPI.py

API will be available at: http://localhost:8000

Test the API:

Interactive docs:

Visit: http://localhost:8000/docs