Skip to main content
Programming turns the tip (or chosen) questionnaire into the target platform deliverable. Progress lives on a survey.programming run (newest-wins per survey). Greenfield runs expose lean status plus workflow_id / stream_url. Clients subscribe on the generic World stream; product resume for target/template uses domain PUT routes.

​
Product steps

Ordered work (anatomy / World steps; UI should follow the live stream rather than inventing a second plan):
  1. Initial programming — questionnaire → Questra HCL (streams compose previews)
  2. Conjoint design — fan-out over cross elements (skipped when none)
  3. Target selection — configure gate; resume via PUT .../programming/target (or generic hook token when present)
  4. Template selection — configure gate; resume via PUT .../programming/template
  5. Transpile — compile HCL once and persist platform source + sidecars
  6. Fine tuning — skipped when effective instructions are empty
  7. Audit / diagnostics loops — bounded repair; exhaustion → warnings
  8. Upload files / survey configuration — API delivery only; skipped for offline targets
Workspace default_file_hosting (questra_cdn vs platform_native) controls whether movable media stay on the Questra CDN or are uploaded upstream. Platforms without a file-manager API always keep movable media on Questra CDN. Offline targets still run through diagnostics and persist local artifacts at every completed stage. Only upstream upload steps skip.

​
Start

202 includes workflow_id, stream_url (/v1/workflows/{workflow_id}/stream), and a lean programming detail (empty phases for greenfield). You may start without a frozen platform target (platform: null). The run waits on configure gates; set target (and optionally template) with the product resume routes below.

​
Subscribe to progress

Prefer the start response stream_url:
Legacy domain subscribe (resolves the current binding for the survey):
NDJSON lines are raw World stream chunks (anatomy + progressive logs). Reconnect with startIndex while the World retains the stream. Authoritative status reconciles from GET /v1/workflows/{workflowId} or GET .../programming.

​
Target and template resume

fine_tuning_template_id: null clears the template and uses the code-owned canonical platform template. Frozen target output includes a secret-free delivery discriminant: { mode: "offline" } or { mode: "api", integration_id, upstream_survey_id? }. Connected resume sends config refs for the integration and optional upstream survey. Export-only freezes mode: "offline". Greenfield anatomy may also expose SDK hook tokens — resume those with POST /v1/workflows/{workflowId}/hooks/{token} when the UI is following resumePath from the stream.

​
Accept / cancel

​
After success

  • GET .../content — tip revision written by programming
  • Webhooks: workflow.*, content.updated
  • Manual PUT .../content is blocked while a programming (or content-edit) run is still pending/running — completed bindings may remain current and do not block saves

​
Concurrency

Newest-wins per survey: a new POST .../programming cancels the prior run. Switch to the new stream_url.