About Zeever.ca
Last updated: April 2026
Zeever.ca is a Canadian AI platform that turns public information into clear, trustworthy answers. We cover all of Toronto.ca, from building permits and property taxes to recycling schedules, parking, recreation programs, and city government.
Founder
Colin Smillie
Founder, Developer, AI Researcher
Colin designs and builds AI systems for Canadian public services, with a focus on retrieval-augmented generation, open-source model evaluation, and Canadian data sovereignty. He publishes research on LLM benchmarking, vector search optimization, and scaling RAG infrastructure.
What we do
Ask a question about any Toronto city service and Zeever.ca will search through official content from Toronto.ca to find what's relevant. It then writes an answer based only on what it found, with citations so you can check the sources yourself.
If the information isn't there, Zeever.ca will say so. It won't guess and it won't fill in the blanks with things it doesn't know.
By the numbers
23,000+
Pages indexed
112,000+
Content chunks
0.94
Relevance score
9
Models tested
Relevance score from 100-question benchmark across 16 categories using 9 open-source models. See our research for full methodology.
Zeever.ca vs. alternatives
| Zeever.ca | Toronto.ca search | Call 311 | Hire a consultant | |
|---|---|---|---|---|
| Speed | Seconds | 10+ minutes | 30+ min wait | Days |
| Cost | Free | Free | Free | $100+/hr |
| Citations | Every answer | You find them | None | Varies |
| Availability | 24/7 | 24/7 | Business hours | By appointment |
| Source | Toronto.ca only | Toronto.ca | City staff | Experience |
How it works
Crawl
We regularly pull in pages and PDFs from across Toronto.ca to keep our content up to date. Smart re-crawling checks for changes so the index stays current.
Understand
That content gets parsed, classified, and broken into focused sections. Each one is turned into a vector embedding that captures what it means.
Retrieve
When you ask a question, we use vector similarity search to find the most relevant sections from across all of Toronto.ca.
Answer
An AI model reads the retrieved evidence and writes a clear answer with citations. It only works from what it found, never from its own training data.
What this is not
Zeever.ca is not a substitute for professional advice. City services can get complicated depending on your situation. For anything important, check with the relevant City of Toronto department directly or talk to a qualified professional.
We're not affiliated with or endorsed by the City of Toronto. We use publicly available content from Toronto.ca to help people find the information they need.
The bigger picture
Toronto.ca is where we started, but the plan is bigger. Zeever.ca is being built as a Canadian AI platform for navigating public services across the country, powered by Canadian infrastructure and open AI models. The goal is to make government processes less painful for everyone.
Technology
Zeever.ca is built using open-source AI models and Canadian-hosted infrastructure wherever possible. We chose not to use OpenAI, Google, or other proprietary AI services. Every language model we use is open source, which means the model weights are publicly available, auditable, and not controlled by a single company.
Right now, we run open-source models through Fireworks.ai for the AI inference layer. Canadian GPU hosting at the scale we need doesn't exist yet, but as that infrastructure comes online we plan to bring inference home too, giving us full data sovereignty from question to answer.
Everything else is already Canadian-hosted: the content database, vector search (PostgreSQL with pgvector), knowledge graph, and the web application itself. And every answer comes with its sources, so you can always verify what you're reading.