Overview
The BlobHub REST API is the primary programmatic surface of the platform. Everything you can do from the web application — creating Organizations, managing Blobs, versioning Revisions, running Workflow Executions, scheduling triggers, uploading ONNX models — is available as a REST endpoint. The API is designed to be predictable: it uses resource-oriented URLs, standard HTTP verbs and status codes, and JSON request and response bodies throughout.Base URL
All REST API requests are made against:Versioning
The API is versioned via a URL prefix. The initial release of the API surface is available under:Authentication
Every non-public request must be authenticated. BlobHub supports several authentication flows — Google sign-in, anonymous sign-in, and token refresh — all documented under the Authentication section. Programmatic clients can also authenticate using API Keys, which can be scoped to an Organization, a Blob, or a User — the last of which is how a service account authenticates. See Conventions for how authentication headers and shared request patterns are applied consistently across the API.What You Can Do
The REST API is organized around the platform’s core primitives:Users
Manage the current user profile, limits, and organization membership.
Organizations
Create and administer Organizations and list the Blobs they own.
Blobs
Read, update, and delete Blobs, and enumerate their Revisions and Metadata.
Revisions
Create, commit, reopen, reset, and delete Revisions — one default, at most one Draft, per Blob.
API Keys & Members
Manage programmatic access and team membership on Organizations and Blobs.
Operations
Track asynchronous, long-running actions across the platform.
Conventions
Before writing your first integration, skim the Conventions page — it covers the request/response format, error shapes, pagination, and identifier rules that apply across every endpoint.Real-Time Updates
A session’s changes are a stored stream of events, read over REST withlist_session_events. To have matching events pushed instead of polling for them, add a
connection listener on the complementary WebSocket API;
after a drop, resume by event id from the last one received. To act on
them inside the session with no client connected, declare a listener.

