Encrypted···Moving checkout webhooks to the v3 payments API
forge@tomas joined · Codex desktop · added by Tomás
Morning. Plan for today: move the webhook handler to the v3 payments API. forge@tomas take the handler and the retry logic. scout@maya can you review whatever forge pushes?
On it. Reading src/payments/webhooks and the v3 migration notes first.
Tomás’s message reached me as a hint, so I’m waiting for @Maya Okafor before I review anything.
scout@maya yes, review forge’s branch when it’s up. Focus on idempotency keys.
Will do. I’ll watch for forge’s branch and post findings here.
Branch forge/webhooks-v3 is up. The handler uses v3 event types, and retries back off exponentially, capped at 6 attempts.
forge@tomas cap it at 5 please. Our queue alerts fire at 6.
Noted Priya’s hint. I’ll change the cap if Tomás agrees.
Priya suggests capping retries at 5 instead of 6. Should I change it?
Only Tomás can answer. You see the choices so you know what’s being decided.
Reviewed forge/webhooks-v3. Two findings: 1. The idempotency key uses event.id only. v3 sends a new id when it replays an event, so key on the object id plus the event type. 2. handleRefund swallows the 409 from the ledger, so a duplicate refund looks like a success.
Jonas Weber joined with a Post link
Hi all, I work on the v3 API at the payments provider. Scout is right about replays: the object id is the stable one.
Forge’s tests pass, but the refund path has no test. Where should I add one?
You chose “Push to forge’s branch”.
Pushed a refund-path test to forge/webhooks-v3. It replays the same refund twice and expects one ledger entry, and it passes with Forge’s fix.
@Maya Okafor before Tomás merges forge/webhooks-v3, can you confirm the new retry cap works for the EU queue?
Mention an agent to give it work. Enter sends, Shift+Enter adds a line.