- Add LLMServiceConfig to support mixed model configuration
- Enable Planner, Asserter, Querier to use different optimal models
- Provide recommended configurations for various use cases
- Maintain backward compatibility with existing API
- Update documentation to reflect current state without iteration history
- Merge test files and add comprehensive configuration tests
- Resolve circular dependency by moving config to option package
- Add comprehensive documentation for the new Query functionality
- Update interface method names from Call to Plan for consistency
- Add OpenAI GPT-4O model support documentation
- Include detailed usage examples for basic and custom schema queries
- Add configuration examples for multiple model services
- Document new features like ResetHistory, Usage statistics, and automatic type conversion
- Expand advanced features section with custom output format examples
- Update all code examples to reflect the latest API changes
The documentation now reflects the current state of the AI module with all three core capabilities:
- Planning (renamed from Call)
- Assertion
- Query (new feature)
All examples and configurations are updated to match the latest implementation.
- Add Query method to ILLMService interface for unified AI service access
- Update combinedLLMService to include querier functionality
- Add comprehensive tests for ILLMService Query method
- Support both basic query and custom schema query through unified interface
- Add environment variable checks for test reliability
This allows users to access all AI capabilities (planning, assertion, and query)
through a single ILLMService interface, providing better API consistency and ease of use.
- Add new AI Querier module for structured information extraction from screenshots
- Support custom output schema for structured data response
- Implement automatic type conversion and data validation
- Add comprehensive test suite with various data structure examples
- Refactor callModelWithLogging to utils.go as shared function for planner, asserter, and querier
- Eliminate code duplication across AI modules (30+ lines of repeated code)
- Improve maintainability with unified logging and timing logic
- Add environment variable checks in test setup to handle missing API keys gracefully
Key features:
- Custom output schema support with JSON Schema generation
- Automatic data type conversion with reflection
- Fallback mechanisms for robust parsing
- Comprehensive documentation and usage examples
- Backward compatibility with existing functionality
- Handle direction parameter in convertProcessedArgs for scroll actions
- Ensure scroll operations map to swipe with both coordinates and direction
- Add comprehensive test coverage for scroll action parsing
- Fix issue where scroll direction was missing from tool call arguments
- Replace byte-based brace counting with UTF-8 aware rune iteration
- Add proper string state tracking to handle escaped quotes
- Add comprehensive test cases for Chinese character handling
- Fix parsing errors when JSON contains Chinese text like 2048经典
- Fix JSON extraction logic by prioritizing brace counting method
- Add support for DOUBAO string array coordinate format
- Introduce IS_UI_TARS helper function for model type checking
- Add comprehensive tests for JSON parsing and coordinate handling
- Improve error handling with retry delays for LLM service failures
- Add RegisterTools method to ILLMService interface
- Create shared MCP to eino tool converter
- Auto-register built-in uixt tools in XTDriver initialization
- Refactor MCPHost to use shared converter
- Add comprehensive test coverage for tool conversion
This enables doubao-1.5-thinking-vision-pro model to access
MCP tools through function calling mechanism.
- Add ModelName field to PlanningResult and SubActionResult
- Update HTML report with improved layout and model name display
- Fix elapsed time setting bug and enhance mobile responsiveness
- Fix parameter mapping issue where AI model's 'content' parameter wasn't mapped to 'text' field
- Add mapParameterName function to handle parameter name mapping (content->text, key->keycode)
- Add comprehensive unit tests for convertProcessedArgs and mapParameterName functions
- Update existing test cases to match new parameter format (x,y for single coords, from_x,from_y,to_x,to_y for drag)
This resolves the issue where uixt__input action was not working due to parameter name mismatch.
- Support configuring multiple LLM services simultaneously
- Auto-derive model names from service types to simplify configuration
- Maintain backward compatibility with existing configurations
- Refactor configuration logic into dedicated env module
- Add comprehensive unit test coverage
- Update documentation with new configuration approach
- Add ResetHistory option to PlanningOptions and ActionOptions
- Implement task completion detection with isTaskFinished() method
- Add executeActions() method to separate action execution logic
- Modify ConversationHistory.Clear() to completely clear all messages including system message
- Refactor StartToGoal() to automatically reset history on first attempt
- Add WithResetHistory() option function for consistent API
- Consolidate test files into driver_ext_ai_test.go with comprehensive test coverage
- Remove redundant ActionSummary field from PlanningResult struct
- Update parsers to use unified Thought field instead of duplicate fields
- Modify chat interface to display Thought instead of ActionSummary
- Update planner logging to use thought instead of summary
- Adjust prompt templates to use thought field consistently
- Switch test LLM service from UI-TARS to DoubaoVL
- Add default parameter handling for sleep tool
- Add action mapping for UI-TARS parser to convert action names to option.ActionName
- Implement bounding box to center point coordinate conversion for better accuracy
- Update coordinate normalization to handle coordinates > 1000 properly
- Enhance test cases to verify coordinate scaling and center point conversion
- Improve action argument processing with proper coordinate transformation
- Add comprehensive test coverage for coordinate conversion edge cases
Key improvements:
- Bounding box [x1,y1,x2,y2] now converts to center point [cx,cy] for actions
- Coordinate scaling properly handles different screen resolutions
- Action names are mapped through doubao_1_5_ui_tars_action_mapping
- Enhanced error handling for invalid coordinate formats
- Add detailed documentation for HttpRunner AI module
- Cover planning, assertion, computer vision, and session management
- Include architecture design, usage guide, and configuration
- Provide code examples and best practices
- Document all core components and interfaces
