●Change8

1.9.0

Breaking Changes
📦 instructorView on GitHub →
⚠ 2 breaking✨ 7 features🐛 6 fixes🔧 9 symbols

Summary

This release introduces Ollama and Writer provider support, improves Gemini and Anthropic integrations, and standardizes VertexAI async parameters. It also enhances error handling with a new exception hierarchy and resolves several dependency conflicts.

⚠️ Breaking Changes

  • The `enable_prompt_caching` parameter has been removed from the Anthropic integration. Users should remove this argument from their function calls as it is no longer supported in the current implementation.
  • Async parameter naming in the VertexAI client has been standardized, which may require updating parameter names in existing VertexAI async calls.

Migration Steps

  1. Remove `enable_prompt_caching` from any Anthropic client calls.
  2. Update VertexAI async calls to match the new standardized parameter naming.
  3. Update dependency constraints if using google-genai or rich.

✨ New Features

  • Improved error handling with a comprehensive exception hierarchy.
  • Added Ollama provider support to auto_client.
  • Implemented JSON mode for the Writer provider.
  • Added Gemini optional support and thought parts filtering in GenAI tool parsing.
  • Enabled Audio module compatibility for Windows.
  • Added dev and docs to project.optional-dependencies for uv compatibility.
  • Updated README and documentation with SEO improvements for asyncio and tenacity.

🐛 Bug Fixes

  • Fixed Gemini configuration issues.
  • Resolved pyright TypedDict key access error in dump_message.
  • Fixed retry mechanism to respect timeout parameters for Ollama compatibility.
  • Handled ThinkingBlock in reask_anthropic_json to prevent parsing errors.
  • Fixed documentation for dynamic model creation examples.
  • Filtered out Gemini thought parts in genai tool parsing.

🔧 Affected Symbols

AnthropicVertexAIOllamaWriterreask_anthropic_jsondump_messageThinkingBlockAudiogenai