understanding-funboost-concepts

Pass a BoosterParams instance to the @boost decorator with queue_name as the only required field. Set task-level options like concurrent_num, qps, max_retry_times, and function_timeout; broker connection details belong in funboost_config.py, not BoosterParams.

push accepts business arguments matching the function signature, while publish takes a dictionary plus TaskOptions for framework controls like task_id, countdown, and eta. Most scenarios only need push.

Funboost imports funboost_config via sys.path order, checking the script directory first then the PYTHONPATH project root. If missing, it auto-generates a template at sys.path[1], so set PYTHONPATH to your project root before running.

Import fct from funboost and access fct.task_id, fct.queue_name, or fct.full_msg inside the consumer function. The fct object is thread-safe and only populated during consumption, unlike Celery's bind=True self pattern.

Consumers run an infinite message-pull loop by design since funboost targets long-running services. During testing, terminate explicitly with a timeout or os._exit(66) rather than expecting automatic shutdown.

Use THREADING for IO-bound tasks, GEVENT or EVENTLET for high-concurrency network work, ASYNC for native async def functions, and SINGLE_THREAD for strict sequential execution. Set it via concurrent_mode in BoosterParams.