Movie Searcher with Redis search types & a customer support RAG implementation

A hands-on workshop to build a movie searcher using Redis Cloud, Vector Search, RAG (Retrieval Augmented Generation), and Semantic Caching. Learn how to implement various search techniques including vector similarity search, hybrid search, full-text search, and more! Also includes a Help Center with guardrails using semantic router and PII protection.

Table of Contents

• Overview
• Workshop Challenges
• Architecture
• Quick Start
• Search Types
• Help Center & RAG
• Troubleshooting
• Resources

---

Overview

This workshop guides you through building a complete movie recommendation system that leverages:

• Redis Cloud - Fully managed Redis database with vector search capabilities
• Vector Similarity Search - Find movies by semantic meaning
• Full-Text Search - Traditional keyword-based search with BM25 scoring
• Hybrid Search - Combine vector and text search for best results
• Filtered Search - Apply metadata filters (genre, rating) to vector results
• Range Queries - Find results within a semantic distance threshold
• Semantic Caching - Cache LLM responses for faster repeated queries
• Help Center RAG - AI-powered customer support with article retrieval
• Semantic Router Guardrails - Block off-topic queries using semantic routing
• PII Protection - Prevent caching of personally identifiable information

🎯 Workshop Challenges

Complete these challenges in order to build out the full application. Look for # TODO comments in the code.

Part 1: Search Engine Fundamentals

#	Challenge	File	Description
1	Index Schema	backend/config.py	Define the Redis index schema with field types (text, tag, numeric, vector)
2	Vector Query	backend/search_engine.py	Create a VectorQuery for semantic similarity search
3	Text Query	backend/search_engine.py	Create a TextQuery for BM25 keyword search
4	Hybrid Query	backend/search_engine.py	Create an AggregateHybridQuery combining vector + text search
5	Range Query	backend/search_engine.py	Modify the distance threshold and observe results

Part 2: Semantic Caching

#	Challenge	File	Description
6	Create Cache	backend/semantic_cache.py	Initialize a SemanticCache with Redis
7	Check Cache	backend/semantic_cache.py	Query the cache for semantically similar entries
8	Store in Cache	backend/semantic_cache.py	Store query-response pairs in the cache

Part 3: Guardrails & Safety

#	Challenge	File	Description
9	PII Detection	backend/guardrails.py	Add regex patterns to detect sensitive information
10	Semantic Router	backend/guardrails.py	Create a SemanticRouter to filter off-topic queries

💡 Tips

• Search for # TODO in your IDE to find all challenges
• Each challenge has hints in the comments above it
• Test your changes using the UI at http://localhost:3000

---

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                         Frontend (React)                         │
│      Movie Search UI  │  Help Center Chat  │  http://localhost:3000
└─────────────────────────────────────────────────────────────────┘
                                │
                                ▼
┌─────────────────────────────────────────────────────────────────┐
│                       NGINX Reverse Proxy                        │
│                     Routes /api/* to backend                     │
└─────────────────────────────────────────────────────────────────┘
                                │
                                ▼
┌─────────────────────────────────────────────────────────────────┐
│                      Backend (FastAPI)                           │
│                   http://localhost:8000                          │
│                                                                  │
│  ┌─────────────────────────────────────────────────────────────┐ │
│  │                    MovieSearchEngine                         │ │
│  │  • Vector Search    • Hybrid Search    • Range Search        │ │
│  │  • Filtered Search  • Keyword Search   • Embeddings Cache    │ │
│  └─────────────────────────────────────────────────────────────┘ │
│                                                                  │
│  ┌─────────────────────────────────────────────────────────────┐ │
│  │                    HelpCenterEngine                          │ │
│  │  • RAG Pipeline     • Semantic Cache   • OpenAI LLM          │ │
│  └─────────────────────────────────────────────────────────────┘ │
│                                │                                 │
│  ┌─────────────────────────────────────────────────────────────┐ │
│  │                      Guardrails                              │ │
│  │  • Semantic Router (topic filtering)                         │ │
│  │  • PII Detection (email, phone, SSN, credit card)            │ │
│  └─────────────────────────────────────────────────────────────┘ │
│                                │                                 │
│                    ┌───────────┴───────────┐                    │
│                    ▼                       ▼                    │
│  ┌─────────────────────────────┐ ┌─────────────────────────────┐ │
│  │   HuggingFace Vectorizer    │ │    OpenAI GPT-4o-mini       │ │
│  │   (all-MiniLM-L6-v2)        │ │    (Response Generation)    │ │
│  └─────────────────────────────┘ └─────────────────────────────┘ │
└─────────────────────────────────────────────────────────────────┘
                                │
                                ▼
┌─────────────────────────────────────────────────────────────────┐
│                          Redis Cloud                             │
│              redis://default:***@your-endpoint:port              │
│                                                                  │
│  ┌────────────────┐ ┌────────────────┐ ┌────────────────────────┐│
│  │  Movie Index   │ │  Help Articles │ │   Semantic Cache       ││
│  │  (HNSW/FLAT)   │ │  Index         │ │   (LLM Responses)      ││
│  └────────────────┘ └────────────────┘ └────────────────────────┘│
│  ┌────────────────┐ ┌────────────────┐                          │
│  │  Embeddings    │ │  Router Index  │                          │
│  │  Cache         │ │  (Guardrails)  │                          │
│  └────────────────┘ └────────────────┘                          │
└─────────────────────────────────────────────────────────────────┘

Quick Start

Prerequisites

• Python 3.11+ & Node.js 20+ (recommended)
• Docker & Docker Compose (alternative)
• Redis Cloud account - redis.io/try-free
• OpenAI API Key - platform.openai.com/api-keys (for Help Center)

1. Redis Cloud Setup

1. Create a free database at redis.io/try-free
2. Copy your connection URL: redis://default:PASSWORD@ENDPOINT:PORT

2. Environment Setup

# Clone and configure
git clone <repo-url>
cd full_stack_redis_ai_workshop

# Create .env file
echo "REDIS_URL=redis://default:YOUR_PASSWORD@YOUR_ENDPOINT:PORT" > .env
echo "OPENAI_API_KEY=your_key_here" >> .env

3. Run Locally (Recommended)

Terminal 1 - Backend:

python3 -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r backend/requirements.txt
uvicorn backend.main:app --reload --port 8000

Terminal 2 - Frontend:

cd frontend
npm install
npm run dev

Access at: http://localhost:5173

4. Import Data & Create Index

# Import movie data
./scripts/import_data.sh

# Create search index (via UI or curl)
curl -X POST http://localhost:8000/api/create-index

Alternative: Run with Docker

docker-compose up --build

Access at: http://localhost:3000

GitHub Codespaces

Make ports 3000, 5173, and 8000 public in the Ports tab.

---

Search Types

Type	Best For	Example Query
Vector	Semantic meaning	"Murder movies with twist"
Keyword	Exact matches	"Murder movies with twist"
Hybrid	Best of both	"college friends story" (alpha=0.5)
Filtered	With constraints	"emotional story" + genre:romance
Range	High relevance only	"smuggling syndicate" + threshold < 0.45

---

Help Center & RAG

The Help Center uses a complete RAG pipeline:

Query → Guardrails → Cache Check → Article Search → LLM Response

Try it:

• "How do I reset my password?" → LLM response
• Same question again → Cached response
• "What's the weather?" → Blocked (off-topic)

Setup:

curl -X POST http://localhost:8000/api/help/ingest

---

Troubleshooting

Connection Issues

If you see connection errors:

1. Verify your Redis Cloud database is running (check the dashboard)
2. Ensure your REDIS_URL is correct in the .env file
3. Check that your IP is whitelisted in Redis Cloud security settings

RIOT Import Issues

If RIOT import fails:

1. Ensure RIOT is installed: riot --version
2. Verify your REDIS_URL environment variable is set correctly
3. Check that the resources/movies.json file exists

Index Creation Issues

If /api/create-index fails:

1. Ensure RIOT import was run first (check for movie:* keys in Redis)
2. Check the backend logs for detailed error messages
3. Verify Redis Cloud connectivity with /api/health
4. Ensure your database has enough memory (30MB free tier should be sufficient)

---

License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Resources

• Redis Cloud - Free Trial
• Redis Vector Library (RedisVL)
• RedisVL Semantic Router
• RedisVL LLM Semantic Cache
• RediSearch Documentation
• RIOT - Redis Input/Output Tools

---