Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.privataswap.com/llms.txt

Use this file to discover all available pages before exploring further.

​
Endpoint

GET /partner/v1/order/{order_id}/stream
Accept: text/event-stream
X-API-Key: pk_live_...
Returns text/event-stream. Heartbeat comment every 15 s. Auto-close on terminal status (completed, failed, expired, refunded).

​
Event format

id: prv-9f0e:confirming:1716293641000
event: order.confirming
data: {"order_id":"prv-9f0e...","status":"confirming","ts":"2026-05-21T10:14:01Z"}
The id is <order_id>:<status>:<unix_ms>. Use it for deduplication if you fan-out to multiple consumers.

​
Resuming after disconnect

Send Last-Event-ID on reconnect. The server has a 60 s in-memory ring buffer per order. Response carries:
HeaderValuesMeaning
Privata-Resume-SourcebufferEvent found in buffer, full replay from there.
Privata-Resume-SourcesnapshotEvent older than buffer (or buffer flushed by restart). Server sends current status, then live.
Privata-Resume-SourcefreshNo Last-Event-ID or first connection.
Privata-Resume-Gap-MsintegerOnly with snapshot: age of last known event in ms.
If you see frequent snapshot resumes, your client disconnects last longer than 60 s. Either fix your network or accept the snapshot fallback — no data is ever lost, only the per-event replay between snapshot and current is.

​
Concurrency limits

TierMax concurrent SSE per partner
Trial50
Starter200
Growth500
Scale1,500
Enterpriseby contract
Above the cap: 429 RATE_LIMITED on connect. One SSE counts as one read in the 30-second-per-read bucket.

​
SDK helper

const stream = privata.streamOrder(orderId);
stream.on("status", (e) => console.log(e.status));
stream.on("resume", (source, gapMs) => console.log("resumed via", source, gapMs));
stream.on("error", (err) => console.error(err));

for await (const event of stream) {
  if (["completed", "failed", "expired", "refunded"].includes(event.status)) {
    stream.close();
    break;
  }
}
The SDK reconnects with Last-Event-ID and exponential backoff (1s, 2s, 5s, 15s, then capped at 30s). After 5 consecutive failed reconnects, it falls back to polling GET /order/{id} every 10 s and emits a degraded event.