Video → Workflow

Purpose: turn a screen recording or demo video into the outline of a MissionSquad workflow — or run general video analysis against your own prompts.

Two capabilities

• Generic video analysis — run a vision‑capable model over a video with your own batch and final prompts (and optional JSON schemas) to get a structured report.
• Video → Workflow — analyze a demo recording and get a structured description of the workflow it demonstrates: the services used, a goal summary, and an ordered list of steps with approximate timestamps. Use this as the starting point to author a workflow config.

Run modes

• Analyze (synchronous): submit and wait for the finished report in the response.
• Runs (asynchronous): submit and get a run id back immediately, then poll or stream progress and read the result when the run completes. Use this for longer videos.

Sources & limits

• Provide a video by file id (a previously uploaded file) or by direct upload (multipart/form-data).
• Uploads are capped at 512 MiB and one file, with common video extensions (.mp4 .mov .mkv .webm .avi, and more).
• Bound the work with options — frame sampling (sampleFps, framesPerBatch, maxTotalFrames), a clip range (clipStartSec/clipEndSec), and token caps.

From report to workflow

The Video → Workflow report is the raw material for a workflow — it is not a ready‑to‑save config. Use its identifiedServices, goalSummary, and steps to author a workflow on the Workflows screen (or via POST /v1/core/workflows).

API parity

• Generic: POST /v1/core/video/analyze, POST /v1/core/video/runs (+ /upload variants)
• Video → Workflow: POST /v1/core/video-to-workflow/analyze, .../runs (+ /upload variants)
• Run lifecycle: GET /v1/core/video/runs/:runId, .../stream, .../cancel

See Video Processing (API) for full request/response types and SSE events.