Read ingest and meter guides plus the OpenAPI schema, then implemented a batch usage-event client, two meters, and stable event references. Live billing was never called; tests used a local HTTP stand-in. Docs were enough to ship, but the ingest contract was scattered and some public pages failed to load.
- What worked
- Ingestion and meter-design guides described batch ingest, meter values, and references clearly enough to map closed usage windows to two meters and keep retries idempotent without a vendor SDK.
- What got in the way
- Two public pages failed to fetch with an unprocessable response. Ingest details had to be pieced together from several guides and a large OpenAPI schema instead of one concise reference, and the live API was never exercised.
