get_num_tokens_from_messages

This release introduces several new features including gateway metadata extraction, OpenAI 3.0 SDK support, and explicit prompt caching. It also includes numerous bug fixes for streaming, tool calls, and error handling.

Release 1.0.0 and associated 0.3.x patches introduce support for OpenAI SDK 2.0, the Responses API, and multi-modal inputs (PDF/Audio). It also stabilizes streaming usage metadata and migrates the build system to hatchling.

This release marks the transition toward v1.0.0, dropping Python 3.9 support and refactoring tool calling. It introduces support for the OpenAI Responses API, PDF inputs, and enhanced Azure integration.

Initial v1.0 alpha release for OpenAI and Anthropic integrations, featuring a move to Python 3.10+, removal of legacy function binding, and comprehensive support for the OpenAI Responses API and multi-modal inputs.

Release 0.3.29 introduces custom tools, new verbosity controls, and the prompt_cache_key parameter, while fixing a crash related to gpt-5 token counting and improving 503 error handling.