Interpretation + bounded planning
Buyer intent · vision · realtime voice / conversation · merchant-interface reasoning
Produces language, observations, candidate choices, and plans—not authority.System architecture
Amana places a deterministic authority layer between AI reasoning, merchant systems, and Razorpay.
AI handles unstructured meaning and planning. Deterministic software controls truth, authority, and money.30-second system view
Buyer intent · vision · realtime voice / conversation · merchant-interface reasoning
Produces language, observations, candidate choices, and plans—not authority.TWO P0 RUNTIME COMMERCE AGENTS
Intent → grounded product → proposal → authorization → payment
Guides the buyer journey; it does not grant authority.STRUCTURED INTENT TO CONTROL PLANEInspect → map → test → evidence → readiness
Proposes mappings and repairs; it does not publish readiness.BOUNDED PLAN TO CONTROL PLANEApplication code
Every material transition is typed, validated, evidence-bound, and fail-closed.
Product grounding · policy evaluation
Material hash · actor-bound approval
Contract evidence · readiness reducer
Checkout; returns provider evidence only
Exact merchant/account · order · amount · INR · captured payment · paid order
Durable outbox · stable refund idempotency · bounded accounting
Authority details
The model may suggest a mapping or repair. Only deterministic logic may publish readiness.
Browser callbacks are evidence, not payment truth.
Operating principle
Model outputs become structured hypotheses, intent, observations, or bounded plans. They are validated before use.
Application code owns grounding, readiness, policy, proposal hashing, authorization, and execution.
Callback, webhook, and provider API evidence is reduced against exact bindings before confirmation.
Explore the system