AI Content Drop Batch and Bulk Operations

An agent acting on many items should not loop one HTTP call at a time. Two endpoints take an array.

Quote many models at once

POST /v1/models/cost takes up to 50 model/quantity pairs and prices them together. Errors are per item, so one unknown model id does not fail the batch.

curl -sX POST https://aicontentdrop.com/v1/models/cost \
  -H "Content-Type: application/json" \
  -d '{"items":[{"model_id":"kling_3_0","quantity":4},{"model_id":"veo_3_fast","quantity":2}]}'

{
  "count": 2,
  "credits_total": 118,
  "items": [
    { "model_id": "kling_3_0", "credits_each": 22, "quantity": 4, "credits_total": 88 },
    { "model_id": "veo_3_fast", "credits_each": 15, "quantity": 2, "credits_total": 30 }
  ]
}

Run several reads in one request

POST /v1/batch takes a list of operations and returns a list of results in the same order. Each entry names a method and a path from the read surface.

curl -sX POST https://aicontentdrop.com/v1/batch \
  -H "Content-Type: application/json" \
  -d '{"operations":[
        {"id":"models","method":"GET","path":"/v1/models?type=image"},
        {"id":"quote","method":"GET","path":"/v1/models/kling_3_0/cost?quantity=3"}
      ]}'

{
  "count": 2,
  "results": [
    { "id": "models", "status": 200, "body": { "count": 12, "models": [] } },
    { "id": "quote",  "status": 200, "body": { "credits_total": 66 } }
  ]
}

Rules, and the reason for each:

• Reads only. Generation is never batched here. A partial failure inside a batch of money-spending calls is ambiguous in exactly the way that costs someone credits they cannot account for; generation gets one request and one Idempotency-Key each.
• Maximum 20 operations per request.
• Per-item status. An operation that fails carries its own status and error; the envelope is still 200.
• No credential required for the read paths it accepts, the same as calling them individually.

In GraphQL

Batching is native: ask for several connections in one query. Batched *transports* (an array of operations in one POST) are refused — the query language already solves that problem. See the GraphQL docs.