llm 0.32.1

LLM 0.32.1 fixes an installation failure caused by relying on another package to bring in a needed dependency. Simon Willison explains that the command-line model tool LLM used the HTTP library httpx, but obtained it indirectly through OpenAI's Python library. When that library stopped depending on httpx, fresh LLM installations broke.

The patch release temporarily restricts the OpenAI package to versions below 3. The failure illustrates why software needs to declare libraries it uses directly: a dependency arriving incidentally through another package can disappear when that package changes.