llm 0.33

This article covers the release of llm 0.33, a command-line tool for accessing large language models. Key highlights include upgrading to the OpenAI Python library 3.x and switching HTTP client dependency to httpx2. New features allow 'llm embed' and 'llm embed-multi' to accept a --key parameter, and embedding methods support per-call keys. Additionally, the 'llm prompt' command now supports repeating -t/--template to combine templates, enabling model configuration and options from one template to be used with another. The release also adds a reasoning_summary option for Responses API models. This is a technical update for developers using the llm tool.