Documentation Index
Fetch the complete documentation index at: https://docs.roughy.ai/llms.txt
Use this file to discover all available pages before exploring further.
What is a pipeline
A pipeline groups one or more processing steps that transform input assets (audio, video) into output assets (cut decisions, rendered video). Pipelines run asynchronously: you POST a definition, get back a pipeline ID, and either poll status or subscribe to a live event stream until the run finishes.Lifecycle
A pipeline moves through five states:| State | Meaning |
|---|---|
PENDING | Created but not yet picked up by a worker. |
RUNNING | At least one step is executing. |
COMPLETED | All steps finished successfully. |
FAILED | A step errored and was not retried. |
CANCELLED | The pipeline was cancelled before it finished. |
- Estimate —
POST /api/v1/pipelines/estimatereturns a price quote plus a signed pricing token (5-minute TTL). - Create —
POST /api/v1/pipelineswith the token reserves credits, creates the pipeline (statePENDING), and queues the first step. - Observe —
GET /api/v1/pipelines/{id}returns the current state plus per-step detail. For real-time updates, subscribe toGET /api/v1/pipelines/{id}/events(Server-Sent Events). - Collect — once
COMPLETED, each step’s output asset is downloadable via the asset endpoints.
Steps
A pipeline contains one or more steps, each binding a processor (cut, render) to a
configuration. Steps execute in declared order; a downstream step
can read the original pipeline input or any prior step’s output,
depending on the processor.
Each step independently moves through PENDING → RUNNING → COMPLETED | FAILED. Step-level state is exposed in the pipeline
detail response.
Cancel and retry
POST /api/v1/pipelines/{id}/cancel— stops aPENDINGorRUNNINGpipeline. Reserved credits are released.POST /api/v1/pipelines/{id}/retry— restarts aFAILEDpipeline from the first step.POST /api/v1/pipelines/{id}/steps/{step_id}/retry— retries a single failed step in place without re-running earlier steps.
Listing
GET /api/v1/pipelines returns a paginated list of your pipelines
(limit, offset query params), most recent first.
Related concepts
- Processor — the individual steps inside a pipeline.
- Asset — the inputs and outputs.
- Credit ledger — how runs are billed.