xby-toronto-open-data-server

Getting a Toronto open-data query answered by an LLM agent used to mean hand-rolling CKAN endpoint calls, guessing dataset IDs, sniffing whether a resource is datastore-backed or a bare CSV, and stitching together filter dictionaries. xby-toronto-open-data-server collapses that pipeline into a small set of semantic tools designed for agent invocation. Start with toronto_search_datasets or toronto_popular_datasets to locate a dataset, then let toronto_smart_data_helper decide the data path, fetch the schema, and return rows or clear next steps in a single call. When precise control is needed, toronto_get_dataset_schema exposes column names and types, and toronto_query_dataset_data accepts filter objects, sort keys, and field projections. For CSV-only resources, toronto_fetch_csv_data pulls headers and sample rows so the agent can inspect structure without a browser.

The workflow follows a simple arc—discover, understand structure, fetch data—and every tool documents its expected inputs and outputs so an LLM can string them together without trial and error. The toronto_start_here tool lays out the recommended sequence on first call, reducing prompt-engineering overhead. API keys are set once via set_xby_apikey and persisted across restarts, so multi-session projects do not require re-authentication.

This plugin fits developers and researchers who build city-data question-answering agents on DeepSeek Harness, need quick access to Toronto traffic, parks, budget, food-inspection, or any of the 500+ CKAN-hosted datasets, and want a single conversational turn to replace a multi-step spreadsheet workflow.