New release of LLM adds support for reasoning traces, OpenAI Responses, server-side tools, and smarter logging
LLM 0.32 turns the CLI and Python API toward traceable, tool-using agent workflows.
The release shows reasoning traces on stderr, adds server-side tools including OpenAI CodeInterpreter and WebSearch, and introduces OpenAI Responses API support. Its Python API now accepts full message lists and can stream typed events for reasoning, text, tool calls, and other outputs. Logging was redesigned around a content-addressable SQLite message store to avoid duplicating full histories on every turn. Willison says the lower-level tool work was driven partly by Datasette Agent, and that LLM is now looking “very agent-shaped” to him. Simon Willison's note
The release shows reasoning traces on stderr, adds server-side tools including OpenAI CodeInterpreter and WebSearch, and introduces OpenAI Responses API support. Its Python API now accepts full message lists and can stream typed events for reasoning, text, tool calls, and other outputs. Logging was redesigned around a content-addressable SQLite message store to avoid duplicating full histories on every turn. Willison says the lower-level tool work was driven partly by Datasette Agent, and that LLM is now looking “very agent-shaped” to him. Simon Willison's note
score 5