llm 0.33
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
The llm library and CLI have upgraded to OpenAI Python SDK 3.x and httpx2, introducing stateless, per-call API key injection for embedding models and collections. This allows concurrent, multi-tenant embedding pipelines to run safely without mutating shared model state. Additionally, new template chaining capabilities enable developers to separate and combine runtime model configurations with prompt templates on the fly.
OpenAI Python library 3.x now enforces httpx2, breaking any production code still pinned to httpx. Re-test all embedding and model calls—key injection is now per-call, so shared state assumptions in plugins or scripts will silently fail unless they’re updated to use the new key= parameter. This also adds repeatable prompt templates, letting you chain model configs with prompts without rewriting your CLI invocations.