conductor(tracks): archive 3 completed tracks, update tracks.md with active/archived sections
This commit is contained in:
112
conductor/archive/comprehensive_gui_ux_20260228/spec.md
Normal file
112
conductor/archive/comprehensive_gui_ux_20260228/spec.md
Normal file
@@ -0,0 +1,112 @@
|
||||
# Track Specification: Comprehensive Conductor & MMA GUI UX
|
||||
|
||||
## Overview
|
||||
|
||||
This track enhances the existing MMA orchestration GUI from its current functional-but-minimal state to a production-quality control surface. The existing implementation already has a working Track Browser, DAG tree visualizer, epic planning flow, approval dialogs, and token usage table. This track focuses on the **gaps**: dedicated tier stream panels, DAG editing, track-scoped discussions, conductor lifecycle GUI forms, cost tracking, and visual polish.
|
||||
|
||||
## Current State Audit (as of 08e003a)
|
||||
|
||||
### Already Implemented (DO NOT re-implement)
|
||||
- **Track Browser table** (`_render_mma_dashboard`, lines 2633-2660): Title, status, progress bar, Load button per track.
|
||||
- **Epic Planning** (`_render_projects_panel`, lines 1968-1983 + `_cb_plan_epic`): Input field + "Plan Epic (Tier 1)" button, background thread orchestration.
|
||||
- **Track Proposal Modal** (`_render_track_proposal_modal`, lines 2146-2173): Shows proposed tracks, Start/Accept/Cancel.
|
||||
- **Step Mode toggle**: Checkbox for "Step Mode (HITL)" with `self.mma_step_mode`.
|
||||
- **Active Track Info**: Description + ticket progress bar.
|
||||
- **Token Usage Table**: Per-tier input/output display in a 3-column ImGui table.
|
||||
- **Tier 1 Strategy Stream**: `mma_streams.get("Tier 1")` rendered as read-only multiline (150px).
|
||||
- **Task DAG Tree** (`_render_ticket_dag_node`, lines 2726-2785): Recursive tree with color-coded status (gray/yellow/green/red/orange), tooltips showing ID/target/description/dependencies/worker-stream, Retry/Skip buttons.
|
||||
- **Spawn Interceptor** (`MMASpawnApprovalDialog`): Editable prompt, context_md, abort capability.
|
||||
- **MMA Step Approval** (`MMAApprovalDialog`): Editable payload, approve/reject.
|
||||
- **Script Confirmation** (`ConfirmDialog`): Editable script, approve/reject.
|
||||
- **Comms History Panel** (`_render_comms_history_panel`, lines 2859-2984).
|
||||
- **Tool Calls Panel** (`_render_tool_calls_panel`, lines 2787-2857).
|
||||
- **Performance Monitor**: FPS, Frame Time, CPU, Input Lag via `perf_monitor`.
|
||||
|
||||
### Gaps to Fill (This Track's Scope)
|
||||
|
||||
1. **Tier Stream Panels**: Only Tier 1 gets a dedicated text box. Tier 2/3/4 streams exist in `mma_streams` dict but have no dedicated UI. Tier 3 output is tooltip-only on DAG nodes. No Tier 2 (Tech Lead) or Tier 4 (QA) visibility at all.
|
||||
2. **DAG Editing**: Can Retry/Skip tickets but cannot reorder, insert, or delete tasks from the GUI.
|
||||
3. **Conductor Lifecycle Forms**: `/conductor:setup` and `/conductor:newTrack` have no GUI equivalents — they're CLI-only. Users must use slash commands or the epic planning flow.
|
||||
4. **Track-Scoped Discussion**: Discussions are global. When a track is active, the discussion panel should optionally isolate to that track's context. `project_manager.load_track_history()` exists but isn't wired to the GUI.
|
||||
5. **Cost Estimation**: Token counts are displayed but not converted to estimated cost per tier or per track.
|
||||
6. **Approval State Indicators**: The dashboard doesn't visually indicate when a spawn/step/tool approval is pending. `pending_mma_spawn_approval`, `pending_mma_step_approval`, `pending_tool_approval` are tracked but not rendered.
|
||||
7. **Track Proposal Editing**: The modal shows proposed tracks read-only. No ability to edit track titles, goals, or remove unwanted tracks before accepting.
|
||||
8. **Stream Scrollability**: Tier 1 stream is a 150px non-scrolling text box. Needs proper scrollable, resizable panels for all tier streams.
|
||||
|
||||
## Goals
|
||||
|
||||
1. **Tier Stream Visibility**: Dedicated, scrollable panels for all 4 tier output streams (Tier 1 Strategy, Tier 2 Tech Lead, Tier 3 Worker, Tier 4 QA) with auto-scroll and copy support.
|
||||
2. **DAG Manipulation**: Add/remove tickets from the active track's DAG via the GUI, with dependency validation.
|
||||
3. **Conductor GUI Forms**: Setup and track creation forms that invoke the same logic as the CLI slash commands.
|
||||
4. **Track-Scoped Discussions**: Switch the discussion panel to track-specific history when a track is active.
|
||||
5. **Cost Tracking**: Per-tier and per-track cost estimation based on model pricing.
|
||||
6. **Approval Indicators**: Clear visual cues (blinking, color changes) when any approval gate is pending.
|
||||
7. **Track Proposal Editing**: Allow editing/removing proposed tracks before acceptance.
|
||||
8. **Polish & Density**: Make the dashboard information-dense and responsive to the MMA engine's state.
|
||||
|
||||
## Functional Requirements
|
||||
|
||||
### Tier Stream Panels
|
||||
- Four collapsible/expandable text regions in the MMA dashboard, one per tier.
|
||||
- Auto-scroll to bottom on new content. Toggle for manual scroll lock.
|
||||
- Each stream populated from `self.mma_streams` keyed by tier prefix.
|
||||
- Tier 3 streams: aggregate all `"Tier 3: T-xxx"` keyed entries, render with ticket ID headers.
|
||||
|
||||
### DAG Editing
|
||||
- "Add Ticket" button: opens an inline form (ID, description, target_file, depends_on dropdown).
|
||||
- "Remove Ticket" button on each DAG node (with confirmation).
|
||||
- Changes must update `self.active_tickets`, rebuild the ConductorEngine's `TrackDAG`, and push state via `_push_state`.
|
||||
|
||||
### Conductor Lifecycle Forms
|
||||
- "Setup Conductor" button that reads `conductor/workflow.md`, `conductor/tech-stack.md`, `conductor/product.md` and displays a readiness summary.
|
||||
- "New Track" form: name, description, type dropdown. Creates the track directory structure under `conductor/tracks/`.
|
||||
|
||||
### Track-Scoped Discussion
|
||||
- When `self.active_track` is set, add a toggle "Track Discussion" that switches to `project_manager.load_track_history(track_id)`.
|
||||
- Saving flushes to the track's history file instead of the project's.
|
||||
|
||||
### Cost Tracking
|
||||
- Model pricing table (configurable or hardcoded initial version).
|
||||
- Compute `cost = (input_tokens / 1M) * input_price + (output_tokens / 1M) * output_price` per tier.
|
||||
- Display as additional column in the existing token usage table.
|
||||
|
||||
### Approval Indicators
|
||||
- When `_pending_mma_spawn` is not None: flash the "MMA Dashboard" tab header or show a blinking indicator.
|
||||
- When `_pending_mma_approval` is not None: similar.
|
||||
- When `_pending_ask_dialog` is True: similar.
|
||||
- Use `imgui.push_style_color` to tint the relevant UI region.
|
||||
|
||||
### Track Proposal Editing
|
||||
- Make track titles and goals editable in the proposal modal.
|
||||
- Add a "Remove" button per proposed track.
|
||||
- Edited data flows back to `self.proposed_tracks` before acceptance.
|
||||
|
||||
## Non-Functional Requirements
|
||||
- **Thread Safety**: All new data mutations from background threads must go through `_pending_gui_tasks`. No direct GUI state writes from non-main threads.
|
||||
- **No New Dependencies**: Use only existing Dear PyGui / imgui-bundle APIs.
|
||||
- **Performance**: New panels must not degrade FPS below 30 under normal operation. Verify via `get_ui_performance`.
|
||||
|
||||
## Architecture Reference
|
||||
- Threading model and `_process_pending_gui_tasks` action catalog: [docs/guide_architecture.md](../../docs/guide_architecture.md)
|
||||
- MMA data structures (Ticket, Track, WorkerContext): [docs/guide_mma.md](../../docs/guide_mma.md)
|
||||
- Hook API for testing: [docs/guide_tools.md](../../docs/guide_tools.md)
|
||||
- Simulation patterns: [docs/guide_simulations.md](../../docs/guide_simulations.md)
|
||||
|
||||
## Functional Requirements (Engine Enhancements)
|
||||
|
||||
### Live Worker Streaming
|
||||
- During `run_worker_lifecycle`, set `ai_client.comms_log_callback` to push intermediate text chunks to the per-ticket stream via the event queue. Currently workers are black boxes until completion — both Claude Code and Gemini CLI stream in real-time. The callback should push `{"text": chunk, "stream_id": "Tier 3 (Worker): {ticket.id}", "status": "streaming..."}` events.
|
||||
|
||||
### Per-Tier Model Configuration
|
||||
- `mma_exec.py:get_model_for_role` is hardcoded. Add a GUI section with `imgui.combo` dropdowns for each tier's model. Persist to `project["mma"]["tier_models"]`. Wire into `ConductorEngine` and `run_worker_lifecycle`.
|
||||
|
||||
### Parallel DAG Execution
|
||||
- `ConductorEngine.run()` executes ready tickets sequentially. DAG-independent tickets should run in parallel via `asyncio.gather`. Constraint: `ai_client._send_lock` serializes all API calls — parallel workers may need separate provider instances or the lock needs to be per-session rather than global. Mark as exploratory.
|
||||
|
||||
### Automatic Retry with Model Escalation
|
||||
- `mma_exec.py` has `--failure-count` for escalation but `ConductorEngine` doesn't use it. When a worker produces BLOCKED, auto-retry with a more capable model (up to 2 retries).
|
||||
|
||||
## Out of Scope
|
||||
- Remote management via web browser.
|
||||
- Visual diagram generation (Dear PyGui node editor for DAG — future track).
|
||||
- Docking/floating multi-viewport layout (requires imgui docking branch investigation — future track).
|
||||
Reference in New Issue
Block a user