This project is a prototype RAG-based parking assistant developed for our partner company OmnyPark. The assistant helps users understand parking charges, free hours, validation rules, and weekend/public holiday parking policies.
It is designed as a public-facing AI chatbot that can be deployed on kiosks, mobile apps, or web platforms to improve user experience and reduce manual queries.
Problem Statement
Parking policies are often confusing for users due to varying rules for:
Free parking duration
Paid parking tariffs
Cinema or retail validation
Weekend and public holiday policies
This project demonstrates how an AI assistant can simplify this by providing quick, clear, and context-aware responses.
Solution Approach
This project uses a Retrieval-Augmented Generation (RAG) architecture:
Parking policy data is stored as structured text files
Data is chunked and embedded using Sentence Transformers
Embeddings are stored in ChromaDB
User queries retrieve relevant context
Google Gemini generates responses based on retrieved context only
This approach ensures:
reduced hallucination
domain-specific answers
better accuracy compared to generic chatbots
Architecture Overview
Knowledge Base ā Text files (data/)
Embedding Model ā Sentence Transformers (MiniLM)
Vector Database ā ChromaDB
LLM ā Google Gemini
UI ā Streamlit
Features
Answer parking-related queries in natural language
Covers tariffs, free hours, validation, and policies
Context-aware responses using RAG
Simple Streamlit UI for interaction
Secure API key handling using .env
Sample Questions
What are the parking charges per hour?
How many free hours are available?
Can cinema visitors get extra parking time?
Is parking free on weekends?
Is parking free on public holidays?
Expected Output
Example:
Question:
What are the parking charges?
Answer:
First 3 hours: Free
After free time: AED 20 per hour
Official site signage remains final
Dataset
The dataset consists of structured text files representing parking policies: