ONNX Celery worker ping flow

[raphy0316]

Summary

Establish the initial Celery infrastructure for the ONNX Runner.
Ensure that the ONNX worker can successfully receive and execute tasks from the Redis queue using a simple onnx.test_ping task and a corresponding debug endpoint.

Why

All subsequent phases of the model execution pipeline (Phase 4–10) depend on a functioning Celery worker system.
Before implementing run logic, adapters, metrics, and storage integration, we must guarantee that:

• Celery can start correctly
• The worker can register tasks
• Redis broker/backend communication works
• API → Redis → Worker → Backend flow is stable

This is a foundational milestone for the entire Runner subsystem.

In Scope

• Initialize ONNX Runner Celery application
• Configure Celery queue name via QUEUE_ONNX
• Implement onnx.test_ping task
• Ensure worker correctly imports tasks (absolute imports)
• Configure worker startup command (celery -A worker -Q ${QUEUE_ONNX})
• Add debug endpoints:
  • /debug/onnx-ping
  • /debug/onnx-ping/{task_id}
• Verify end-to-end flow using test task

Out of Scope

• run_experiment() implementation (Phase 4)
• DRY_RUN lifecycle logic (Phase 5)
• ONNX adapter or runtime execution (Phase 8–10)
• Metrics implementation (Phase 6)

Acceptance Criteria

• ONNX worker container successfully loads Celery app without errors
• Redis broker/backend connections succeed
• Calling /debug/onnx-ping returns a valid task_id
• Worker logs show receipt and execution of the test task:
  • [ONNX-RUNNER] test_ping called ...
• /debug/onnx-ping/{task_id} returns:
  • state: SUCCESS
  • result containing the echoed payload

Tasks

• Implement onnx.test_ping task (task.py)
• Update worker.py to import task module using absolute import (import task)
• Fix worker command to load Celery app properly (celery -A worker worker -l info -Q ${QUEUE_ONNX})
• Add /debug/onnx-ping endpoint to dispatch test task
• Add /debug/onnx-ping/{task_id} endpoint to inspect task result
• Validate worker logs for task execution
• Add developer documentation for Celery + Redis flow

Labels

• backend
• celery
• runner
• phase-1
• feature

Assignees

• @raphy0316