Introduction

Understanding the different types of Blobs available.

BlobHub supports various specialized Blob Types, each tailored with data storage and processing engines for specific use cases. Every blob carries exactly one type: a dotted string of the shape <namespace>.<content kind>.<name>, chosen when the blob is created and fixed for its lifetime. The internal data structure, storage semantics, and API commands for a blob depend entirely on that string.

Five type strings exist today:

Type string	Blob Type	Description
blobhub.compute.workflow	Workflow	Backend workflows: definitions, sessions, executions, events.
blobhub.compute.scheduler	Scheduler	Schedules that fire workflow executions on a cadence.
blobhub.model.onnx	ONNX	Hosting for ONNX machine learning models.
blobhub.graph.orientdb	Graph	General-purpose graph storage in a dedicated OrientDB database.
blobhub.dataset.nlp	NLP Dataset	NLP dataset content.

An organization can create a type only while the type’s enabled entitlement is true for it. workflow.enabled, scheduler.enabled and onnx.enabled are true by default; orientdb.enabled and nlp.enabled are false. See Limits for every limit of each type.

Workflow

The Workflow Blob Type (blobhub.compute.workflow) is designed to build, manage, and execute complex backend workflows.

• You can treat memory and data as first-class citizens with built-in state management.
• Supports managing Definitions, initiating and monitoring Executions within Sessions, and accessing real-time Execution Events.

Scheduler

The Scheduler Blob Type (blobhub.compute.scheduler) stores a collection of schedules that trigger workflow executions on a time-based cadence.

• Supports both one-time and recurring cron schedules with full timezone control.
• Retains a history of past executions, including success and failure details for each firing, for 30 days by default.

ONNX

The ONNX Blob Type (blobhub.model.onnx) provides specialized hosting for Open Neural Network Exchange (ONNX) machine learning models.

• Upload a model archive in one request, or in parts over presigned URLs when it is larger — see Upload Model and Multipart Upload.
• Describe the artifacts the platform derives from an upload, and download each one either in the response itself (Download Model) or, for a larger one, in ranged parts over presigned URLs (Multipart Download).
• The Python SDK wraps all of it.

Graph

The Graph Blob Type (blobhub.graph.orientdb) backs general-purpose graph storage in a dedicated OrientDB database. Not documented in depth here; the Python SDK runs OrientDB SQL against it.

NLP Dataset

The NLP Dataset Blob Type (blobhub.dataset.nlp) stores NLP dataset content. It has no data-plane commands of its own — a data/command or data/query call against it answers unsupported_operation — so it is created and read through the platform’s shared primitives rather than through type-specific commands. Not documented in depth here.