POST /v1/systemone/batch
Send many requests in one call.
POST https://api.openinstinct.dev/v1/systemone/batchUse a batch to label a list of items, or to ask about several states at once. The requests are spread over the model's replicas and answered in order.
Request body
| Field | Type | |
|---|---|---|
requests | array | 1 to 256 requests, each of the shape /v1/systemone takes |
- All requests of a batch must use the same model (or all leave
modelout). - The whole body is limited to 32 MB. With images, size the batch accordingly.
curl https://api.openinstinct.dev/v1/systemone/batch \
-H "Authorization: Bearer $OPENINSTINCT_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"requests": [
{
"state": "Where is my invoice for March?",
"questions": { "is_billing": { "type": "noul", "instructions": "Is this about billing?" } }
},
{
"state": "The app crashes when I open settings.",
"questions": { "is_billing": { "type": "noul", "instructions": "Is this about billing?" } }
}
]
}'Response
200 OK, with one entry per request, in the order they were sent:
{
"responses": [
{
"model": "instinct-one-latest",
"answers": { "is_billing": { "type": "noul", "noul": 0.97 } },
"usage": { "input_tokens": 24, "output_tokens": 17 },
"latency_ms": 41
},
{
"error": { "status": 422, "detail": "..." }
}
],
"count": 2,
"errors": 1
}| Field | |
|---|---|
responses | For each request, a /v1/systemone response or {"error": {"status", "detail"}} |
count | The number of entries |
errors | How many of them failed |
A batch returns 200 even when some of its requests failed. Check errors, or each entry for an error field.
Limits and cost
- Each request of a batch counts toward the requests per minute limit; the batch as a whole takes one concurrent slot. See Rate limits.
- Usage is the sum over the answered requests.