Versionhq

- Refined multimodal handling [audio-text]
- Refined CUA handling using Playwright and selenium.
- Minor error handling

- Added Azure AI and Azure OpenAI as custom endpoint providers
- Refined credential and endpoint handling of LLM class
- Minor fix on task/tool handling
- Refined structured response handling for json object
- Added o3 series
- Updated documentation

- Streamlined dependencies
- Refined Gemini (thinking exp) handling
- Aggregate scores
- Refined step callback handling
- Refined OpenAI Responses API handling

- Added OpenAI's Response API handling
- Added OpenAI's tools - File Search, Web Search, and CUA
- Streamlined task execution params and logics
- Added Web/File Search results as prompt context
- Refined utility functions
- Updated docs

- Enhanced agent class methods
- Improved response schema validation and handling *Terminated output type classifications
- Integrated Gemini 2.0 features and constants *Terminated legacy models
- Updated docs

- Refined usage tracking with token consumption, latency, and error tracking
- Updated docs