Smartrentals AI REST API- Technical Architecture Report

Author : Shriniwas Kulkarni

• PCCOE 2026 BTech CSE(AIML)
• Email: kshriniwas180205@gmail.com
• Phone: +91 [8999883480]
• GitHub: github.com/Shriniwas18K

Overview

This is a sophisticated property recommendation system built using FastAPI, leveraging advanced technologies for secure, scalable, and intelligent property matching.

System Architecture

Key Technical Components

1. Backend Framework

• FastAPI: High-performance, modern Python web framework
• Enables rapid API development with automatic OpenAPI (Swagger) documentation
• Implements robust request validation using Pydantic models

2. Authentication Mechanism

• Token-Based Authentication
  • Custom token generation using Fernet encryption
  • 10-minute session validity
  • Secure token validation process
• Prevents multiple simultaneous logins
• Cryptographically secure token generation

3. Database Integration

• PostgreSQL: Relational database for user management
• Stores user credentials and transaction logs
• Tables:
  • credentials: User registration details
  • transactions: User activity tracking

4. Vector-Based Recommendation System

• Pinecone Vector Database: Enables semantic property search
• Sentence Transformers: Converts property addresses to high-dimensional embeddings
• Advanced recommendation algorithm using cosine similarity
• Supports intelligent, context-aware property recommendations

5. Security Features

• CORS Middleware: Configurable cross-origin resource sharing
• Environment variable management with python-dotenv
• Parameterized database queries to prevent SQL injection
• Cryptographic token generation

Technical Highlights

Advanced Search Capability

def get_recommendations(pinecone_index, search_term, top_k=10):
    embed = get_embeddings([search_term])
    res = pinecone_index.query(vector=embed, top_k=top_k, include_metadata=True)
    return res

• Converts search terms into vector embeddings
• Retrieves semantically similar properties
• Supports flexible, intelligent search

Robust Data Validation

class Property(BaseModel):
    PropertyTypes: Literal['1 Bedroom', '2 Bedroom', ...]
    Security: Literal['Not Applicable', 'Gated Community', ...]
    # ... other strictly typed fields

• Uses Pydantic for type enforcement
• Ensures data integrity
• Supports predefined value sets for specific fields

Technology Stack

• Web Framework: FastAPI
• Database: PostgreSQL
• Vector DB: Pinecone
• ML Model: Sentence Transformers
• Encryption: Cryptography (Fernet)
• ORM/Database Driver: Psycopg2

Scalability and Performance Considerations

• Serverless Pinecone vector index
• Efficient embedding generation
• Stateless API design
• Minimal external dependencies

Potential Improvements

• Implement more advanced authentication (e.g., JWT)
• Add rate limiting
• Enhance error handling
• Implement more sophisticated recommendation algorithms

Conclusion

A modern, scalable property recommendation system demonstrating expertise in:

• API design
• Machine learning integration
• Database management
• Security implementation