●Change8

1.10.0

Breaking Changes
📦 instructorView on GitHub →
⚠ 2 breaking✨ 7 features🐛 5 fixes⚡ 1 deprecations🔧 11 symbols

Summary

This release introduces native caching (Redis/AutoCache), expands provider support to include DeepSeek and Anthropic parallel tools, and migrates Google integrations to the new google-genai SDK.

⚠️ Breaking Changes

  • Integrations have been updated to use the from_provider API, which may require updating how clients are initialized.
  • The google-generativeai provider is being phased out in favor of google-genai.

Migration Steps

  1. Update provider integrations to use the new from_provider API.
  2. Switch from google-generativeai to the google-genai package for Gemini models.
  3. Update provider syntax in codebases to match the latest documentation standards.

✨ New Features

  • Added native caching support with AutoCache and RedisCache adapters.
  • Enhanced Batch API with multi-provider support and improved CLI.
  • Added DeepSeek provider support.
  • Added Anthropic parallel tool support.
  • Added comprehensive api_key parameter support for all providers in auto_client.
  • Added Bedrock-native format conversion to OpenAI format.
  • Added automatic Partial wrapping for streaming requests in GenAI.

🐛 Bug Fixes

  • Fixed batch request parsing issues.
  • Fixed GenAI modes to handle response_model=None.
  • Corrected is_simple_type logic for list types containing BaseModel contents.
  • Fixed GenAI to forward thinking_config parameters to Gemini models.
  • Fixed Decimal type support for GenAI provider.

🔧 Affected Symbols

from_providerAutoCacheRedisCachegoogle-genaigoogle-generativeaiis_simple_typethinking_configBedrockAnthropicDeepSeekauto_client

⚡ Deprecations

  • google-generativeai is deprecated; use google-genai instead.