Remove Agent Desk users, roles, login sessions, tokens, and local permission persistence. Expose the backend as an embeddable ai-agent module with host-provided subject lookup and operation authorization callbacks, and complete the frontend/backend repository split.
- Removed runtime mode handling from AIAgentConfigWorkbench and related components.
- Updated tests to reflect changes in AI Agent policy copy and configuration.
- Changed terminology from "workflow" to "revision" in various components and API responses.
- Simplified agent binding logic in channel editing.
- Cleaned up unused variables and types related to runtime modes.
- Updated localization files for consistency with new terminology.
- Deleted AgentRunLogResponse struct from skill_response.go.
- Removed agent_run_log_repository.go and agent_run_log_service.go files.
- Eliminated AgentRunLogDetailDialog component and its related logic from the dashboard.
- Removed agent run log fetching functions from admin API.
- Updated navigation and localization files to remove references to agent run logs.
- Updated import paths in multiple service files to reflect the new agent-desk module.
- Added a new configuration file for agent-desk in the Docker setup.
- Updated runtime configuration to use __CS_AI_AGENT_WIDGET_CONFIG__ instead of __CS_AGENT_WIDGET_CONFIG__.
- Changed message types in support host bridge from "cs-agent" to "cs-ai-agent".
- Minified SDK script updated to reflect new AI agent naming conventions.
- Adjusted scrollbar styles in main.scss to use .cs-ai-agent-scrollbar instead of .cs-agent-scrollbar.
- Moved the skill routing logic from matcher.go to a new router.go file for better organization.
- Implemented replyCommitService to handle sending AI replies and managing reply rounds.
- Added replyInterruptService to manage conversation interrupts and resume handling.
- Created replyRunLogService to log AI reply actions and their outcomes.
- Introduced helper functions for building conversation interrupts and resolving prompts.
- Added unit tests for the new services and functions to ensure correctness.
- Removed unused code and optimized imports in matcher.go.
- Added AllowedToolCodes field to Context for managing tool access.
- Introduced ResumeSource in aiReplyTraceData to track resume points.
- Enhanced logging with additional fields: PlannedSkillName, SkillRouteTrace, InterruptType, ResumeSource, and FinalStatus.
- Implemented functions to parse and resolve allowed tool codes for agents and skills.
- Refactored skill definition creation and update logic to streamline request handling.
- Updated MCP tool catalog to include source type and integrated built-in tools.
- Improved UI components to display additional tool information and enhance user experience.
t