API documentation

sinio.ai turns the raw events your systems already emit — badge reads, work orders, sensor readings, tickets — into episodes, habits and findings, without asking you to model any of that yourself. You push facts; the platform derives meaning.

Getting started

Create an API key from your project's settings in the sinio.ai console, then send your first batch to POST /v1/events. There is nothing to configure first: the reference is never a condition of acceptance, so an event pointing at a device sinio.ai has never heard of is still accepted, and the device is created for you as undeclared. You enrich the reference afterwards, at your own pace.

Four things worth knowing up front

• Ingestion is asynchronous. A 200 confirms that the batch was received, not that it was derived. Derivation happens on its own queue, moments later.
• Ingestion is idempotent on (project, source, source_event_id, revision). Replaying a call is always safe — a replay is counted in duplicates, never stored twice.
• Acceptance is partial. Within one batch, valid events are accepted even when others are rejected. Always read rejections.
• No response should ever lose an event silently. A 503 is explicitly replayable; a 4xx is something you fix rather than retry.