Simon Willison Releases llm 0.33 with OpenAI 3.x and httpx2 Upgrades
Simon Willison has released llm 0.33, upgrading the tool's dependencies to the OpenAI Python library 3.x and switching its HTTP client to httpx2. The update also introduces the ability to pass keys directly to embedding commands and combine multiple prompt templates sequentially. This release improves the flexibility of the llm CLI tool by allowing users to chain prompt templates and manage API keys dynamically for embedding models. It also ensures compatibility with modern HTTP clients and the latest OpenAI library standards. The new --key option for embedding commands passes the resolved key to plugins without altering the shared model state. Additionally, reasoning-capable models now support a reasoning_summary option with auto, concise, and detailed settings.
## BACKGROUND
llm is a popular command-line tool and Python library developed by Simon Willison that allows users to interact with various Large Language Models (LLMs) both locally and via remote APIs. Vector embeddings are high-dimensional representations of text used in machine learning to capture semantic meaning. Prior to this release, managing API keys for embedding models differed from how they were handled for regular LLM prompts.