Single session-level registry of every analysis (one-shot + recurring) across all entry kinds.
Results live here, and per-entry views read them on demand via Analyses.getAnalysesForEntry.
analyseData registers a record and a job on JobEngine. monitorAnalysis toggles the job's
active flag.
The displayed base map's style, language, viewport and layers.
Bring-your-own-data GeoJSON layers, customer-authored data the agent can read and render.
Strategy used to run code-generation tools (analyseData, processData, createTracker,
clusterIncidents). createToolState initialises it to the env-resolved executor, and
createMapAgent re-resolves it with MapAgentOptions.codeExecution. Tools read this
to select where sandbox code runs.
Derived/custom polygon entries that processData outputs.
The single recurrence engine. Owners (analyses, trackers) register entry-anchored jobs and keep
the returned handles. The engine re-runs each active job when its source entries change. It is generic
and never interprets a job's recipe (the tools layer builds the run closure). See JobEngine.
Map POI layer visibility, filtering and category management.
The places layer's PlacesModule and append-only search history.
Reachable range results, with origin, budgets and bbox summaries.
Route calculation, waypoint management, and planning parameter state.
Generic trackers, rising-edge reducers fed by their own JobEngine jobs. A tracker fires an
alert/event when its rule's Verdict crosses an edge (ingestVerdict). The slice
keeps the durable TrackerEvent log.
Traffic area analytics aggregation module, last result and viz config.
Fetched traffic incident entry history and per-entry rendering.
Traffic flow tiles and incident overlay tiles.
State passed to tool factory functions, organised by feature area. Each sub-state mixes lazy module access with the current service data produced during the session.
Example