Graph Engine v1

Thailand decision engine.

This is where the graph turns signal rows into traveler-facing answers about where to go, where to stay, which route is fragile, and what to check before paying.

6 active families Explainable travel answers Decision API

What this layer does.

Decisions are the interface layer over entities, observations, and signals. They explain the recommendation, the weak spot, the next move, the pivot, and the score components instead of hiding the answer inside a black box.

Today

Pick the strongest place now.

Destination decisions rank Bangkok, Chiang Mai, Phuket, and the rest against live comfort and confidence signals.

Stay and move

Choose the base and route shape.

Hotel-area, route-check, and ferry-check decisions sit between raw data and payment decisions.

Why trust it

Weakness stays visible.

Every decision carries source confidence, a weak signal, and the checks you should still run before paying.

Decision family registry.

These are the traveler-facing decision families the graph can compute today.

Active decision family

Where to go today

Ranks destination entities using live AQI and comfort signals plus source confidence, regional context, and airport-scale, airport-access, airport board-surface, or airport environmental-support context where it materially changes the trip shape.

Decision type: where_to_go_today

Output: Best and weakest destination picks today.

Active decision family

Where to stay

Ranks hotel-area entities using comfort, source confidence, parent-destination resilience, airport-ground-access support, official airport pickup-rule context, airport board-surface support, airport-scale context, airport environmental-support context, rail-support context, and official rail presence where they materially affect the stay.

Decision type: where_to_stay

Output: Best current hotel bases and who they fit.

Active decision family

Route check

Checks high-value route templates using destination scores, transfer nodes, airport-weather context, airport-ground-access support, official airport pickup-rule context, official passenger board-surface support, airport-scale context, airport environmental-support context, rail-support context, official rail presence where the route can plausibly use it, and live pier or transfer support where coastal legs matter.

Decision type: route_check

Output: Route feasibility, main breaker, and pivot.

Active decision family

Ferry check

Checks pier and island-chain plans with live sea-state caution, public transfer support, ferry fragility, and honest refund logic.

Decision type: ferry_check

Output: Ferry fragility, refund logic, and backup move.

Active decision family

Food route

Turns destination and venue-density entity records into traveler-facing food-route picks.

Decision type: food_route

Output: Best food-first destination and route fit.

Active decision family

Gold buying check

Explains the national Thai Gold Traders Association 96.5% reference quote in THB per 1 baht-weight, then pairs it with city buying-context friction so Bangkok, Chiang Mai, Phuket, Pattaya, and Hat Yai are not mistaken for separate official local price feeds.

Decision type: gold_buying_check

Output: National quote, city buying context, and next gold-shopping checks.

Live decision status.

The browser hydrates this from /api/decisions. It shows the top current outputs from each family, plus the strongest next actions and pivots.

Registry surfaces.

Changed since last run

Browser refresh compares the latest persisted decision snapshots here.

Live decision dashboard

Generated by the live decision API. The browser hydrates current decision outputs here.

Each decision row now exposes component scores, penalties, and fallback markers so the engine can be audited instead of only trusted.

Live decision status updates in the browser from the decision API.

The static registry already knows about 6 decision families. Browser refresh fills this panel from the API with live ranked answers.