Back to the archive

The Promise Ledger / End-to-end verified

Commerce that remembers the promise

A buyer-side runtime that turns purchase intent into verifiable promises, follows fulfillment, and resolves breaches through policy-gated payments and refunds.

PythonFastAPINext.jsTypeScriptPostgreSQLRazorpay
* CHECK
SUBJECT
DANTE
TYPE
Agentic commerce
STATUS
End-to-end verified

* 197 / 197 deterministic evaluation cases passed.

PROOF SIGNAL

197 / 197deterministic evaluation cases passed

The mission

What the system does

A buyer-side runtime that turns purchase intent into verifiable promises, follows fulfillment, and resolves breaches through policy-gated payments and refunds.

  • Compiles buyer intent into typed constraints and evaluates offers deterministically
  • Freezes evidence-backed merchant promises into a hashed purchase contract
  • Builds a purchase-rights graph connecting promises, evidence, breaches, and remedies
  • Verifies signed webhooks and executes idempotent, policy-approved test-mode refunds

Engineering log

Decisions behind the build

What the system handles, and why it works this way.

BOSS 01

Who controls the money?

AI can propose an action, but deterministic services validate the contract, amount, authorization, and policy before any payment or refund executes.

BOSS 02

Events arrive twice

Duplicate and out-of-order webhooks reconcile into the same purchase state. Raw-body HMAC verification and idempotency keys protect the execution boundary.

BOSS 03

Promises need proof

Material promises retain typed values, source evidence, and a contract hash so a later remedy can be traced back to the buyer's original intent.

System map

How signals move

A plain technical view. The map still works without hovering.

01
Buyer briefNext.js

Intent, offers, and consent

02
Promise ledgerFastAPI

Typed, hashed contracts

03
Payment gateRazorpay

Signed test-mode events

04
Rights graphDomain engine

Evidence and remedies

05
Audit recordPostgreSQL

Durable event history

LOG ENTRY / WHAT I LEARNED

“An agent can suggest what should happen. Financial authority needs a smaller, deterministic system that can explain exactly why it allowed it.”

Source signal

Inspect the work.