Ontario Building Code Chat Assistant

An interactive AI-powered chat interface for navigating and understanding the Ontario Building Code (v25). This application uses advanced natural language processing and vector search to provide accurate, context-aware responses to questions about building regulations in Ontario.

🌟 Features

• Interactive Chat Interface: Modern Streamlit-based UI with real-time response streaming
• Smart Search:
  • Query expansion for improved search relevance
  • Vector-based semantic search using ChromaDB
  • Progress tracking for search operations
• Context-Aware Responses: Utilizes gpt-4o-mini for generating accurate, contextual answers
• Comprehensive Token Tracking:
  • Total processed tokens across all operations
  • Conversation-only tokens (excluding RAG context)
  • Document context tokens from RAG retrieval
  • Separate input/output token tracking
  • Real-time cost estimation based on token types
• Flexible Authentication:
  • Password-protected access with app password
  • OpenAI API key authentication for custom billing
  • Session persistence for both methods
• Source References: Displays relevant sections from the building code
• User-Friendly Features:
  • Expandable search query display
  • Progress bars for operations
  • Word count for retrieved sections
  • Chat history management
  • Token usage and cost tracking

🚀 Getting Started

Prerequisites

• Python 3.8 or higher
• UV package manager

Installation

1. Clone the repository:

git clone https://github.com/kmaurinjones/Ontario-Building-Code-Chat.git
cd Ontario-Building-Code-Chat

2. Create and activate a virtual environment:

python -m venv .venv
source .venv/bin/activate  # On Windows, use `.venv\Scripts\activate`

3. Install dependencies using UV:

uv pip install -r requirements.txt

4. Set up environment variables:

cp .env.example .env

Edit .env with your configuration:

• OPENAI_API_KEY: Your OpenAI API key (default key for app password users)
• APP_PASSWORD: Password for accessing the application

Running the Application

Start the Streamlit app:

uv run streamlit run app.py

The application will be available at http://localhost:8501 by default.

🔒 Security

• Dual authentication system:
  • App password authentication (uses default API key)
  • OpenAI API key authentication (uses user's API key)
• Session-based authentication
• Secure password hashing
• Persistent authentication across page reloads
• Dynamic API key management
• API key validation

📚 Data Source

Currently references Ontario Building Code v25 from: https://www.ontario.ca/laws/regulation/120332/v25

🛠️ Project Structure

Ontario-Building-Code-Chat/
├── app.py              # Main Streamlit application
├── src/               # Source code modules
├── data/              # Vector database and data storage
├── setup/             # Setup and initialization scripts
├── requirements.txt   # Project dependencies
└── .env              # Environment variables

⚠️ Disclaimer

This tool is designed to assist in navigating the Ontario Building Code. Always verify information with the official building code documentation for critical decisions.

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

🔄 Changelog

See changelog.md for a detailed history of changes and updates.

📊 Token Counting System

The application implements a comprehensive token counting system that tracks various types of tokens:

Token Types

• Total Processed Tokens: All tokens processed by AI throughout the session
• Conversation Tokens: Tokens in the cleaned conversation history (system prompt + chat)
• Document Context Tokens: Tokens from RAG context retrieval
• Input/Output Tokens: Separate tracking for model input and output

Token Counting Process

1. Query Expansion:

   • Counts tokens in conversation history + query (input)
   • Counts tokens in expanded queries (output)

2. RAG Context:

   • Counts tokens in retrieved document chunks
   • Tracks separately from conversation tokens

3. Chat Completion:

   • Counts tokens in full prompt with context (input)
   • Counts tokens in model response (output)
   • Updates conversation tokens without RAG context

Cost Calculation

• Input tokens: $0.15 per 1M tokens
• Output tokens: $0.60 per 1M tokens
• Real-time cost estimation in sidebar