StepHandler abstraction (app/Support/VideoWorkflow/StepHandlers/) plugs
into WorkflowEngine without coupling it to how any given step type is
actually executed:
- ScriptGenerationHandler: builds a prompt from the project's title/
description (or an explicit outline/tone/keywords/target length in
step input), generates via an active OpenAI AiProvider, stores the
result as a 'script' asset.
- TtsVoiceoverHandler: voices step.input.text, or falls back to the
project's latest script asset, via an active ElevenLabs AiProvider;
voice_id/model_id are configurable per step. No direct "speed"
control — ElevenLabs' base TTS endpoint doesn't expose one.
RunWorkflowStepJob resolves a step's handler by type and runs it on
the queue, matching the existing PublishPostJob pattern (including a
stale-dispatch guard: no-op if the step isn't 'in_progress' anymore).
UI, under its own /video-projects routes and nav entry — separate
from /projects per the earlier decision:
- VideoProjects/Index (#43): list projects with status and step
progress, create new ones.
- VideoProjects/Show (#44): start a project from a WorkflowTemplate or
a hardcoded default 8-step pipeline; per-step Run/Approve/Reject/
Retry/Skip actions; auto-queues the next step after any transition
when a handler exists for its type (steps with no handler yet just
sit in_progress rather than auto-failing); lists project assets.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
