🚗 Car Listing Submitting Tool

AI system that converts unstructured car listing text and images into standardized JSON data for automotive platforms, with automated email delivery to designated recipients.

⚡ What It Does

Converts unstructured car listings (text + image) → structured JSON → email delivery to designated recipients.

🎯 Core Features

• 🧠 AI-Powered Extraction - GPT-4 with hardened prompts (injection-resistant)
• 🔒 Strict Validation - Pydantic models with year ranges (1975-2025)
• 📧 Email Delivery - Sends processed listings to designated recipients
• 🛡️ Robust Implementation - Token limits, error handling, logging

🚀 Quick Start

git clone https://github.com/Youssef-Ashraf-Dev/Car_Listing.git
cd car-listing-processor
pip install -r requirements.txt
cp .env.example .env  # Add your Azure OpenAI + Gmail credentials
streamlit run app.py

Tech Stack: Python • Streamlit • LangChain • Azure OpenAI • Pydantic

📋 Sample Input/Output

Input Text:

Blue Ford Fusion produced in 2015 featuring a 2.0-liter engine. The vehicle has low mileage with
only 40,000 miles on the odometer. Equipped with brand-new all-season tires manufactured in 2022.
The car's windows are tinted for added privacy. Notably, the rear bumper has been replaced after
a minor collision. Priced at 1 million L.E.

Input Image:

Output JSON:

{
    "car": {
      "body_type": "sedan",
      "color": "Blue",
      "brand": "Ford",
      "model": "Fusion",
      "manufactured_year": 2015,
      "motor_size_cc": 2000,
      "tires": {
        "type": "brand-new",
        "manufactured_year": 2022
      },
      "windows": "tinted",
      "notices": [
        {
          "type": "collision",
          "description": "The rear bumper has been replaced after a minor collision."
        }
      ],
      "price": {
        "amount": 1000000,
        "currency": "L.E"
      }
    }
}

🏗️ Solution Architecture

Key Files:

• app.py - Streamlit UI & orchestration
• llm_parser.py - Hardened prompts + Pydantic models
• utils.py - Email delivery + image analysis
• .env.example - Configuration template

Architecture Diagram

Sequence Diagram

🛡️ Security & Hardening

Prompt Injection Protection:

• Strict system instructions that forbid responding to user commands
• Schema-locked output (JSON only, no markdown/comments)
• Temperature 0 for deterministic results

Data Validation:

• Year constraints (1975-2025 for cars, 2000-2025 for tires)
• Required field validation with clear error messages
• Optional field omission (no hallucinated data)

Error Handling:

• Specific catches for API, parsing, and SMTP failures
• Comprehensive logging for debugging
• Development fallback: saves to local outbox/ directory when email credentials are missing (testing convenience)

🧪 Testing & Validation

# Test individual components
python llm_parser.py       # LLM parsing test
python utils.py            # Email delivery test (needs test_car.jpg)

# Run full workflow via UI
streamlit run app.py

🎛️ Production Optimizations

• Token Efficiency: Optimized prompts reduce costs
• Input Limits: 800 char descriptions, 5MB images
• Capped Responses: 300 tokens max (sufficient for complete JSON)