- Add VCR patch for binary request bodies (base64 encoding fallback)
- Add generate_filename() utility for UUID-based filenames with extension
- Add OpenAIResponsesFormatter for Responses API (input_image, input_file)
- Fix OpenAI uploader to use 'vision' purpose for images
- Fix Anthropic uploader to use tuple format (filename, content, content_type)
- Add TextConstraints and text support for Gemini
- Add file_id upload integration tests for Anthropic and OpenAI Responses API
- Add get_file_uploader() method to BaseLLM (returns None by default)
- Implement get_file_uploader() in Anthropic, OpenAI, Gemini, Bedrock
- Pass both sync and async clients where applicable
- Update uploaders to accept optional pre-instantiated clients
- Update factory to pass through client parameters
This allows reusing authenticated LLM clients for file uploads,
avoiding redundant connections.
- Make crewai_files an optional dependency with graceful fallbacks
- Move file formatting from executor to LLM layer (_process_message_files)
- Add files field to LLMMessage type for cleaner message passing
- Add cache_control to Anthropic content blocks for prompt caching
- Clean up formatters: static methods for OpenAI/Gemini, proper error handling
- Remove unused ContentFormatter protocol
- Move test fixtures to lib/crewai-files/tests/fixtures
- Add Azure and Bedrock multimodal integration tests
- Fix mypy errors in crew_agent_executor.py
