manual_slop/conductor/tracks/comprehensive_gui_ux_20260228/spec.md

Track Specification: Comprehensive Conductor & MMA GUI UX

Overview

This track aims to implement a comprehensive, high-density, and tactile GUI for controlling all aspects of the Conductor orchestration and the 4-Tier Hierarchical Multi-Model Architecture (MMA). The focus is on providing the user with explicit, granular control over the entire lifecycle of a track, from initialization to task execution, with a "Manual First" philosophy.

Goals

1. Conductor Lifecycle Management: Implement UI components for /conductor:setup, /conductor:newTrack, and a robust Track Browser.
2. MMA Orchestration Dashboard: Create a multi-viewport modular system for monitoring and managing the 4-Tier hierarchy.
3. Task DAG Management: Implement an interactive DAG visualizer that allows the user to reorder, insert, and delete tasks/tickets incrementally.
4. Tier-Specific Discussion History: Provide dedicated visual panels for each tier agent's (Orchestrator, Tech Lead, Worker, QA) discussion history and streaming output.
5. Manual Execution Control: Ensure every event, task transition, and sub-agent spawn requires manual user acceptance by default, with opt-in automation.
6. Metrics & Observability: Integrate real-time telemetry, token usage, and cost tracking for each tier and track.

Functional Requirements

• Track Browser: A central hub for viewing, selecting, and managing implementation tracks.
• Interactive Task DAG: A node-based UI (using Dear PyGui) for visualizing task dependencies. Users can manually manipulate the graph (reorder, add, remove nodes).
• Tier History Panels: Separate viewing modes for the logs/discussions of each tier agent.
• Execution Clutch: A global toggle for "Step Mode" vs. "Auto-Queue", with the default being manual step-through.
• Spawn Interceptor: An improved modal for reviewing and editing worker prompts/context before they are dispatched.
• Observability Widgets: High-density monitors for FPS, CPU, and API metrics (latency, token burn, cost).

Non-Functional Requirements

• Arcade Aesthetics: Use blinking notifications, visceral color-coded feedback (Ready, Running, Blocked, Done), and high-contrast visuals (USA Graphics style).
• Asynchronous Responsibility: The UI must remain responsive during multi-model generation using the AsyncEventQueue.
• Modular Layout: Support for docking, floating, and rearranging all Conductor/MMA windows.

Acceptance Criteria

• Users can initialize and manage tracks entirely through the GUI.
• The Task DAG accurately reflects the state of the active track and allows manual editing.
• Each tier agent's history is viewable in a dedicated panel.
• The "Manual Acceptance" flow is enforced for all MMA actions by default.
• API metrics and telemetry are visible and update in real-time.

Out of Scope

• Automated "Auto-Fix" loops without user intervention.
• Remote management via web browser (staying local GUI for now).