Manpreet Kaur

Posted on Aug 11

RediFlow AI

#redischallenge #devchallenge #database #ai

Redis AI Challenge: Real-Time AI Innovators

This is a submission for the Redis AI Challenge: Real-Time AI Innovators.

What I Built

RediFlow AI is an intelligent data processing and chatbot platform that combines Redis 8's vector storage capabilities with AI-powered natural language interfaces. The project integrates a Redis MCP (Model Context Protocol) Server with an advanced product recommendation system for seamless data ingestion, vector search, and intelligent query processing.

Core Features:

Advanced Product Recommendation Engine: AI-powered product recommendations based on feature preferences with dynamic slider-based tuning
Vector-Powered Semantic Search: High-performance similarity search using Redis 8's HNSW vector indexing
Intelligent Chatbot: LLM-powered query analysis with automatic MCP tool selection
JSON Data Processing: Intelligent processing and indexing of complex product catalog data
Feature-Based Preference Learning: Interactive system for adjusting product feature importance (0.0-1.0 scale)
MCP Integration: Full Model Context Protocol support for AI agent workflows

Demo

🎥 Video Demo: Watch Video

🚀 Live Demo: Demo link

Demo Username: admin
Demo Password: password123

📁 GitHub Repository: GitHub link

📚 Documentation: Complete Setup Guide

Application Workflow:

Upload Data: Upload JSON product catalogs to Redis vector database
Chat Interface: Ask natural language questions about stored products
Recommendation System:
- Search for products using conversational queries
- Adjust feature preferences with interactive sliders
- Get personalized product recommendations ranked by compatibility score

How I Used Redis 8

RediFlow AI leverages Redis 8's cutting-edge vector search capabilities as its core real-time data layer, implementing several advanced AI-focused features:

🔍 Vector Similarity Search with HNSW

Implementation: Created Redis 8 vector indices using the HNSW (Hierarchical Navigable Small World) algorithm for approximate nearest neighbor search
Code Example:

# Create vector index with HNSW algorithm
index_def = IndexDefinition(prefix=["doc:"])
schema = VectorField(
    "vector",
    "HNSW", 
    {"TYPE": "FLOAT32", "DIM": 1536, "DISTANCE_METRIC": "COSINE"}
)
r.ft("vector_index").create_index([schema], definition=index_def)

🧠 Semantic Product Search

Feature: Implemented semantic search using OpenAI embeddings (1536 dimensions) stored as Redis 8 vectors
Benefit: Users can find products using natural language queries like "printers with good performance" instead of exact keyword matching
Performance: KNN search with configurable similarity thresholds for relevance filtering

📊 Real-Time Recommendation Engine

Vector Storage: Product features and metadata stored in Redis hashes with associated vectors
Similarity Matching: Advanced algorithms to find products matching user preferences using vector distance calculations
Dynamic Scoring: Real-time calculation of product compatibility scores based on feature similarity

🎛️ Interactive Feature-Based Filtering

Implementation: Combined Redis vector search with feature-based preference learning
User Experience: Dynamic sliders allow users to adjust importance of product features (battery_life, camera_quality, performance, etc.)
Algorithm: Calculates weighted similarity scores between user preferences and product feature vectors

⚡ High-Performance Data Operations

Multi-Data Structure Support: Leveraged Redis hashes, JSON documents, lists, sets, and sorted sets alongside vectors
Real-Time Processing: Instant product recommendations and search results
Scalable Architecture: Designed to handle large product catalogs with efficient vector indexing

🔧 Model Context Protocol (MCP) Integration

Innovation: Built a comprehensive Redis MCP server exposing 40+ Redis operations as AI tools
AI Agent Compatibility: Seamless integration with Claude Desktop, VS Code Copilot, and Cursor
Natural Language Interface: LLM automatically selects appropriate Redis operations based on user queries

🚀 Production-Ready Features

Connection Management: Redis Cloud integration with SSL support
Error Handling: Comprehensive error handling and connection resilience
Monitoring: Built-in database statistics and performance monitoring
Authentication: Secure user authentication and session management

Technical Architecture: The system processes JSON product data, generates embeddings using Azure OpenAI, stores vectors in Redis 8 with HNSW indexing, and provides real-time semantic search and recommendations through a Streamlit interface powered by the Redis MCP server.

Key Redis 8 Features Utilized:

Vector Search Engine: HNSW algorithm for approximate nearest neighbor search
COSINE Distance Metric: Optimal for semantic similarity calculations
Float32 Vector Storage: Efficient storage of 1536-dimensional embeddings
Index Management: Dynamic index creation and management
Hybrid Data Storage: Combination of vectors, hashes, and JSON documents
Real-Time Query Processing: Sub-millisecond search responses
Scalable Vector Operations: Handles large-scale product catalogs efficiently

Architecture Overview

┌─────────────────────────────────────────────────────────────┐
│                    RediFlow AI Platform                     │
├─────────────────────────────────────────────────────────────┤
│  Intelligent Chat & Recommendation System (Streamlit)      │
│  ├── User Authentication & Session Management              │
│  ├── JSON Data Processing & Vector Indexing               │
│  ├── Natural Language Query Processing                     │
│  ├── Advanced Product Recommendation Engine               │
│  └── Feature-Based Preference Learning                     │
├─────────────────────────────────────────────────────────────┤
│  Redis MCP Server                                          │
│  ├── Model Context Protocol Implementation                 │
│  ├── Vector Similarity Search Engine                       │
│  ├── Multi-Data Structure Management                       │
│  └── Natural Language Interface                            │
├─────────────────────────────────────────────────────────────┤
│  Redis 8 Vector Database                                   │
│  ├── HNSW Vector Indexing & Search                        │
│  ├── Product Metadata & Feature Storage (Hash)            │
│  ├── Real-time Messaging (Pub/Sub)                        │
│  └── Event Streaming & Processing                          │
└─────────────────────────────────────────────────────────────┘

Impact & Use Cases

Business Applications:

E-commerce Platforms: Personalized product recommendations based on user preferences
Retail Analytics: Understanding customer preferences through interaction data
Content Discovery: AI-powered content recommendation systems
Customer Service: Intelligent product support and automated recommendations

Technical Innovation:

MCP Protocol: First comprehensive Redis MCP server implementation
Hybrid Search: Combining semantic vector search with feature-based filtering
Real-Time AI: Sub-second response times for complex recommendation queries
Scalable Architecture: Production-ready system handling large product catalogs

Developed by: ShorthillsAI Team

Managers: Siddhartha Jain Kapil Saxena
Developers Team:
- Manpreet Kaur (Dev-Username: manpreetshorthillsai)
- Rajsee Panwar (Dev-Username: rajsee_panwar_cd470873c4b)
- Piyush (Dev-Username: piyush_dtu)

Thank you for this incredible challenge opportunity! The Redis MCP Server has enabled us to build something truly revolutionary in the search and analytics space. 🚀

⚠️ By submitting this entry, you agree to receive communications from Redis regarding products, services, events, and special offers. You can unsubscribe at any time. Your information will be handled in accordance with Redis's Privacy Policy.

DEV Community