E-Commerce Search Engine (MVP)

A search and ranking microservice for an e-commerce platform focused on electronics, optimized for Tier-2 and Tier-3 cities in India.

The system supports typo-tolerant search, Hinglish queries, price-sensitive ranking, and business-aware relevance scoring.

---

Problem Statement

In a large electronics catalog (millions of products), users often search using:

• Typos (e.g. Ifone 16)
• Hinglish queries (e.g. Sasta iPhone)
• Price constraints (e.g. iPhone 50k rupees)
• Attribute intent (e.g. color, storage, accessories)

The goal is to return relevant and affordable products, ranked intelligently and within <1000ms latency.

---

Architecture

• FastAPI microservice
• In-memory catalog (for low latency)
• Modular layers:
  • API layer (app/api)
  • Business logic (app/services)
  • Models (app/models)
• Easily extensible to Elasticsearch / datastore later

---

Ranking Strategy (Core Design)

The ranking algorithm is hybrid and explainable:

1. Text relevance

   • Fuzzy matching (RapidFuzz)
   • Handles typos and partial matches

2. Query intent

   • Hinglish normalization (sasta, sastha)
   • Price intent (50k, rupees)
   • Accessory intent (cover, charger)
   • Freshness (latest)

3. Business signals

   • Product rating
   • Units sold
   • Stock availability

4. Price fitness

   • Boosts affordable products for Tier-2/3 users

This ensures affordable, relevant, and available products rank higher.

---

APIs Implemented

1. Store Product

POST /api/v1/product

2. Update Product Metadata

PUT /api/v1/product/meta-data

3. Search Products

GET /api/v1/search/product?query=...

All APIs are documented via Swagger.

---

How to Run Locally

pip install -r requirements.txt
python -m uvicorn app.main:app --reload