// the find
ydf0509/funboost
pip install funboost，python全功能分布式函数调度框架,funboost的功能是全面性重量级，用户能想得到的功能99%全都有;funboost的使用方式是轻量级，只有@boost一行代码需要写。支持python所有类型的并发模式和一切知名消息队列中间件，支持如 celery dramatiq等框架整体作为funboost中间件，python函数加速器，框架包罗万象，用户能想到的控制功能全都有。一统编程思维，兼容50% python业务场景，适用范围广。只需要一行代码即可分布式执行python一切函数，99%用过funboost的pythoner 感受是 简易 方便 自由 强大，相见恨晚 。
funboost is a Python distributed task scheduling framework built around a single @boost decorator that turns any function into a distributed consumer. It supports an unusually wide range of brokers — Redis, RabbitMQ, Kafka, Postgres, SQLite, TCP sockets, and more — and handles concurrency mode selection, QPS throttling, retries, and deduplication as decorator parameters. It targets Python developers who want Celery-like capability without restructuring their code around a framework's opinions.
The dual-run design is genuinely useful: calling fun(x, y) runs the function directly, fun.push(x, y) sends it to the queue. You can develop and test without a broker running, then flip to distributed execution without changing the function body. The broker abstraction is real — the codebase has actual consumer and publisher implementations for 40+ backends, not just a config flag that breaks at runtime. Stacking multiprocessing on top of threading or asyncio via a single parameter is a clean solution to GIL limitations that would otherwise require writing IPC code by hand. The BoosterParams inheritance pattern for shared config across related tasks is a practical ergonomic win over Celery's per-task decorator proliferation.
The performance claims (22x publish throughput, 46x consume throughput over Celery) are not substantiated in the README — the benchmark methodology is opaque and the numbers are implausible without a specific controlled scenario. With 40+ broker implementations, a FaaS layer, CDC binlog support, OpenTelemetry tracing, Prometheus metrics, workflow orchestration, a web management console, APScheduler integration, and remote deployment tooling all maintained by what appears to be a single developer, the quality of less-used integrations is a real concern — some of these are almost certainly thin wrappers that break on non-trivial use. The PYTHONPATH dependency is a footgun in production deployments, and the README explicitly warns about it without a good resolution. Documentation is primarily in Chinese with English available only through translation artifacts, which makes evaluating edge-case behavior difficult for anyone not reading Chinese fluently.