TL;DR: This release fixes a bug where image blocks in Anthropic tool results were silently dropped during conversion to OpenAI format, breaking multimodal tool outputs.
Summary: The fix ensures that when a tool_result contains image blocks, they are properly converted to OpenAI multimodal content parts (text + image_url array). Plain-text results remain unchanged. This preserves multimodal capabilities for tools that return images.
Why it matters: For AI builders using multimodal agents with tools, this ensures reliable conversion between API formats, preventing silent failures. Developers integrating Anthropic and OpenAI tool outputs should update to this version to maintain image-based tool feedback.
Source: github_releases