tanitmap-ai

What it is

tanitmap-ai is the first agentic map for Tunisian agricultural data. A researcher, journalist, or analyst types a natural-language question; an AI agent decides which data to fetch through agridata MCP, retrieves the answer live from agridata.tn, and renders it adaptively as a chart, a choropleth map of Tunisia, a sortable table, or a short text — depending on the shape of what came back. Every answer carries the source dataset, the last update date, and a direct link to verify on the portal.

Where agridata MCP exposes Tunisia’s open agricultural data to any LLM client through a Model Context Protocol server, tanitmap-ai is the visual front door: it removes the install step (no client setup, no local server config) and turns the studio’s data infrastructure into a one-page interface a user can open in a browser.

How it works

A query input takes a natural-language question. An agent decides which agridata MCP tools to call to retrieve the answer, fetches the data live from the agridata.tn portal, and selects a rendering format based on the shape of the result. Every answer carries the exact dataset name, last update date, license, and a direct link to verify on the portal.

Source attribution is non-negotiable. A response without a link back to agridata.tn is treated as a failure case in the agent loop, not as an acceptable output.

What’s available today

A static demo, available at tanitdata.org/tanitmap-ai/, replays six pre-recorded answers captured from real backend runs against agridata.tn. Real data, real source attributions, real timing — the demo replays them faithfully so you can see what the live tool will feel like before its full version ships.

The demo’s six guided answers cover representative question types: production by governorate, yield over time, comparison of regions, geographic concentration of a crop, climate context, and a multi-step synthesis. Each one walks through the rendering: a chart, a map, a table, or a short text answer, with the source bullet visible throughout.

What’s coming

The full LLM-backed version, where any question is answered live by the agent rather than replayed from a fixture, is the next milestone. It will ship on the dedicated subdomain tanitmap-ai.tanitdata.org once the agent loop is stable enough for open-ended queries on the studio’s public infrastructure. No timeline is committed publicly until the operational dependencies are in place.

Status and roadmap

tanitmap-ai is in active development. The demo is a stable preview; the live version follows the same agent loop on the same data sources, swapping pre-recorded fixtures for live MCP calls. Updates will accompany the project’s milestones — additional demo questions in the short term, the live tool when ready.