AI Image Workflow

This is the runtime pipeline your users interact with directly when they click "Generate".

Feature Page Screenshot

End-to-End Flow in Bunship

1. Frontend calls POST /v1/ai/generations.
2. Task is queued via the pluggable task queue (Trigger.dev / BullMQ).
3. The adapter dispatches the task to the configured backend.
4. Provider returns output, then files are uploaded to object storage.
5. Wallet credits are consumed/refunded based on task outcome.

Core APIs

• Create task: POST /v1/ai/generations
• Query task: GET /v1/ai/generations/:taskId
• Cancel task: POST /v1/ai/generations/:taskId/cancel
• Retry task: POST /v1/ai/generations/:taskId/retry
• User task list: GET /v1/ai/generations

Code Map

• Route layer: apps/ship-api/src/module/ai/generations.ts
• Task queue adapters: apps/ship-api/src/services/ai/queue/ (details)
• Shared processor: apps/ship-api/src/services/ai/queue/processor.ts
• Image app page: apps/ship/src/app/[locale]/(marketing)/ai/image

What Buyers Usually Change

1. Default models and prompt presets in UI.
2. Timeout/retry/concurrency strategy for your budget.
3. Error messages and user-facing task state copy.
4. Credit pricing and free-tier limits.

Launch Checklist

1. Queue backlog and worker concurrency are monitored.
2. Failures trigger refunds/retries as expected.
3. Generated files are accessible by valid URLs.
4. High-cost model usage is protected with strict limits.

Next Steps