LLM 0.32.1 pins OpenAI Python below 3.0 to avoid extra httpx install

According to the maintainer's release note, the pin is intended to keep fresh LLM installations working without a separate installation of httpx. The change is narrowly about the package environment: it controls which major version of the OpenAI Python library LLM may resolve during a new install.

This is not presented as a new model capability or a broad change to LLM's command-line workflow. It follows LLM 0.32, which was released earlier in August, and addresses the installation path for the subsequent patch release.

Dependency constraints can be consequential for developer tools because a clean installation is often where a new major dependency requirement first appears. By retaining an OpenAI Python version below 3.0.0, the patch aims to avoid requiring users to separately add httpx merely to complete a fresh LLM install.

That does not establish compatibility for every existing environment. Projects that pin LLM or the OpenAI Python client directly should review their lockfiles and run their normal installation and smoke-test steps before adopting the patch. The source record supports the dependency reason, not a blanket compatibility promise.

GitHub lists 0.32.1 as a release from Aug. 21, and PyPI records the same version and date. PyPI's provenance details link the package artifact to the same simonw/llm tag, giving the release identity and publication date an independent registry record alongside the maintainer's release note.