npc0/TorontoOpenDataQA

Toronto Open Data QA [WIP, contribution needed]

AI-powered question answering system for Toronto's open datasets.

Features

• —🔍 Smart Search: Searches through 500+ Toronto open datasets
• —🤖 AI Analysis: Uses Claude to understand your question and select relevant data
• —📊 SQL Generation: Automatically generates and executes SQL queries
• —✅ Full Transparency: Shows all steps, queries, and results for verification
• —📝 Natural Language: Get answers in clear, easy-to-understand format

How to Use

1. 1.Get a free API key from Anthropic Console
2. 2.Enter your question about Toronto data
3. 3.Paste your API key
4. 4.Click "Ask"

The system will:

• —Search the Toronto Open Data index
• —Select the most relevant datasets
• —Download and convert the data
• —Generate SQL queries to answer your question
• —Provide a comprehensive answer with sources

Example Questions

• —"What are the top 5 most dangerous intersections in Toronto?"
• —"How has parking ticket revenue changed over the years?"
• —"Which neighborhoods have the most tree coverage?"
• —"What are the most common types of 311 service requests?"

Privacy

• —Your API key is used only for the current session
• —No data is stored or logged
• —All processing happens in real-time
• —Query results are shown for full transparency

Technical Details

Stack:

• —Gradio for UI
• —DuckDB for SQL queries
• —Anthropic Claude for AI
• —Docling for file conversion

Data Source: Toronto Open Data Index

Verification

All answers include:

• —Links to source datasets
• —SQL queries used
• —Number of rows analyzed
• —Steps to manually verify results

Limitations

• —Depends on Docling API availability (may have rate limits)
• —Not all file formats are supported
• —Data is as current as the Toronto Open Data portal

Contribution

Known issues on the download and import dataset from open data portal to duckdb not stable, conversion fails a lot, needs further inspection and debugging.

License

MIT License