Objects are registered records inside a bucket that reference one or more blobs (files or JSON). Objects are validated against the bucket schema. Creating an object does not execute processing; collections handle processing later.

​
Overview

  • Object: Logical container you create in a bucket; holds one or more blobs and optional metadata.
  • Blob: A single file or JSON payload referenced by an object; must match the bucket schema field type.
  • Lineage: Downstream documents keep lineage to each object and blob.
  • Processing: Use batches or collection pipelines to process after registration.

​
Object shape

Object creation uses this request body shape. See the full endpoint at Create Object. 
{
  "key_prefix": "/products/red-sneaker",
  "metadata": {"category": "shoes", "year": 2024},
  "skip_duplicates": true,
  "blobs": [
    {
      "property": "image_main",
      "type": "image",
      "data": "https://example.com/images/red-sneaker-front.jpg",
      "metadata": {"mimetype": "image/jpeg"}
    },
    {
      "property": "specs",
      "type": "json",
      "data": {"title": "Red Sneaker", "sizes": [7,8,9,10]}
    }
  ]
}
  • key_prefix (optional): Logical path prefix applied to the object; helps organize downstream documents.
  • metadata (optional): Arbitrary JSON. Propagates to documents produced by collections.
  • skip_duplicates (optional): When true, identical blobs (by content hash) are skipped.
  • blobs (required): One or more blob entries.

​
Blob shape

Blobs map to fields in your bucket schema. See field types in your bucket definition. 
{
  "property": "image_main",
  "key_prefix": "/images", 
  "type": "image",
  "data": "https://example.com/images/red-sneaker-front.jpg",
  "metadata": {"mimetype": "image/jpeg"}
}
  • property (required): Schema field name in the bucket.
  • key_prefix (optional): Path segment for this blob (appended to the object’s prefix if provided).
  • type (required): Must match the schema field type (e.g., image, video, pdf, json).
  • data (required): The payload. For files, a URL; for JSON fields, inline JSON.
  • metadata (optional): Blob-level metadata; merged into downstream processing context.

​
Batch create

Register multiple objects in a single request using Create Objects in Batch. 
{
  "objects": [
    { "key_prefix": "/products/red-sneaker", "blobs": [ {"property": "image_main", "type": "image", "data": "https://example.com/red.jpg"} ] },
    { "key_prefix": "/products/blue-sneaker", "blobs": [ {"property": "image_main", "type": "image", "data": "https://example.com/blue.jpg"} ] }
  ]
}

​
Behavior & validation

  • Schema enforcement: Each blob’s property and type must match the bucket schema; invalid objects are rejected.
  • Prefixes: object.key_prefix and blob.key_prefix combine to form a logical path carried into downstream documents.
  • Metadata propagation: object.metadata is copied to downstream documents (collections may add more).
  • Idempotence: Use skip_duplicates to avoid re-registering the same content.

​
See also

Uploads

How to submit single and batch object registrations

Buckets

Define schemas and organize objects

Collections

Configure feature extractors and outputs

Batching

Submit grouped objects for asynchronous processing