Cloudflare Releases Python Workers to Public

Cloudflare announced the general availability of Python Workers two years after the initial preview, positioning Python alongside JavaScript and TypeScript on the Developer Platform. This update integrates with Workers AI, R2, D1, Hyperdrive, Durable Objects, Queues, and Workflows, while adding compatibility with FastAPI, Django, and Flask.

The general availability of Python Workers was made possible by three key pieces of engineering, two of which were developed outside of Cloudflare. The first is a packaging standard, which allows Python packages with C, C++, or Rust extensions to be cross-compiled to WebAssembly. This standard, known as PEP 783, was proposed by Cloudflare and accepted after more than a year.

Engineering Developments

The packaging standard enables maintainers to build their own wheels using the Pyodide build toolchain and cibuildwheel, which now supports PyEmscripten. This development removes Cloudflare’s dependency on compiling and hosting packages itself, which previously limited the number of available packages.

The second key engineering development is a socket bridge, which enables drivers such as aiomysql and asyncpg to work unmodified in a WebAssembly sandbox. Cloudflare implemented the necessary system calls over the Workers connect API, allowing for seamless integration with Hyperdrive.

The third development involves upstream contributions to HTTP clients, including requests and httpx, which now route through the JavaScript fetch API in WebAssembly environments. This change enables OpenAI, LangChain, and MCP to run inside a Worker.

The bindings themselves have also been improved, with type conversion now happening inside the runtime and the SDK. Web frameworks can be used through connectors, with workers.asgi and workers.wsgi translating incoming JavaScript requests into the structures expected by WSGI and ASGI applications.

Community Reaction

The community reaction to the general availability of Python Workers has been sharply divided, with some commentators expressing concerns about the underlying engineering and maintenance. One commentator, illia-v, a maintainer of the urllib3 project, noted that while his project had merged contributions adding Pyodide and Emscripten support, the funding for these contributions went to the contributor rather than the maintainers.

This has led to concerns about the security and maintenance of the upstream pieces, with illia-v pointing out that the Emscripten backend remains experimental in urllib3 and is explicitly out of scope in the project’s security policy. He also warned of potential differences in behavior between the browser networking semantics and urllib3’s usual backend.

Other commentators, such as Syrus Akbary, founder of Wasmer, have raised concerns about cold starts and performance, with Akbary noting that a minimal Python application started in around 60ms on Wasmer Edge compared to around 900ms on Cloudflare Workers. Dominik Picheta, from Cloudflare, responded that the company’s memory snapshot implementation has improved cold starts significantly, and that sharding reduces the frequency of cold starts.

Akbary also raised concerns about version coupling, with Cloudflare’s workerd being tied to a specific version of Python and Pyodide. Picheta noted that compatibility flags can select between different Python versions, but Akbary argued that this still creates a problem, as the flag moves workerd too.

Commenter dangoodmanUT raised a separate concern about memory usage, noting that the changes are likely to eat into the worker memory allocation. Hood Chatham, a Pyodide contributor, responded to a technical point raised by Akbary, arguing that patching Python’s event loop to use the JavaScript loop does not create incompatibility, as Python coroutines stay lazy.

Simon Willison suggested that Cloudflare send money toward Pyodide, and Picheta noted that two of the authors of the announcement, Gyeongjae Choi and Hood Chatham, are Pyodide core developers or contributors. Cloudflare has published production patterns in a python-workers-examples repository and updated its developer documentation to show Python alongside JavaScript and TypeScript across products.

Cloudflare says it plans to make Python Workers more performant and memory-efficient, but for now, the company has not released specific performance figures. The python-workers-examples repository provides examples of how to use Python Workers in production, and the developer documentation offers guidance on getting started with Python on the Cloudflare platform.