STEP 01
Typed graph traversal
A GROQ query walks a bounded, typed reference graph in Sanity to enumerate every candidate award routing — no free-text guessing, just the documents and the edges between them.
Proves the cheapest award‑travel routing. Won’t guess when sources disagree.
An agent that walks a typed Sanity knowledge graph, pauses the moment its sources contradict each other, and hands the arithmetic to a deterministic solver that shows its proof.
Every routing, chart, and contradiction here is a synthetic dataset built to exercise the reasoning — not real award pricing. The discipline is the point.
How it works
STEP 02
Contradiction pauses pricing
The traversal hits an engineered contradiction — a printed award chart says 85k, a devaluation notice says 90k. Pricing stops. Nothing is priced until the disagreement is resolved against the authoritative source.
STEP 03
Deterministic proof
A deterministic TypeScript solver returns the single cheapest valid routing and a minimality proof: every other valid routing, priced higher. The model does no arithmetic.
The moment it refuses to guess
Two sources disagree on the same award. Below is a static preview of what the solver surfaces — the live tools render the real thing.
SOURCE A · AWARD CHART
85,000
SFO → NRT business · printed chart
SOURCE B · DEVALUATION NOTICE
90,000
SFO → NRT business · updated notice
Pricing pauses here. The agent resolves the conflict against the authoritative source before the deterministic solver returns a single proven routing — it never averages the two or picks one at random.