ed/manual_slop
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

more refinements

Browse Source
This commit is contained in:
Edward R. Gonzalez
2026-03-06 15:47:18 -05:00
parent fca40fd8da
commit 49ae811be9
2 changed files with 149 additions and 106 deletions
Download Patch File Download Diff File Expand all files Collapse all files

125
conductor/tracks/cost_token_analytics_20260306/spec.md
View File

@@ -8,59 +8,48 @@ Real-time cost tracking panel displaying cost per model, session totals, and bre
### Already Implemented (DO NOT re-implement)
#### cost_tracker.py (src/cost_tracker.py)
- **`MODEL_PRICING` dict**: Pricing per 1M tokens for all supported models
- **`MODEL_PRICING` list**: List of (regex_pattern, rates_dict) tuples
```python
MODEL_PRICING: dict[str, dict[str, float]] = {
"gemini-2.5-flash-lite": {"input": 0.075, "output": 0.30},
"gemini-2.5-flash": {"input": 0.15, "output": 0.60},
"gemini-3.1-pro-preview": {"input": 1.25, "output": 5.00},
"claude-3-5-sonnet": {"input": 3.00, "output": 15.00},
"deepseek-v3": {"input": 0.27, "output": 1.10},
# ... more models
}
MODEL_PRICING = [
(r"gemini-2\.5-flash-lite", {"input_per_mtok": 0.075, "output_per_mtok": 0.30}),
(r"gemini-2\.5-flash", {"input_per_mtok": 0.15, "output_per_mtok": 0.60}),
(r"gemini-3-flash-preview", {"input_per_mtok": 0.15, "output_per_mtok": 0.60}),
(r"gemini-3\.1-pro-preview", {"input_per_mtok": 3.50, "output_per_mtok": 10.50}),
(r"claude-.*-sonnet", {"input_per_mtok": 3.0, "output_per_mtok": 15.0}),
(r"deepseek-v3", {"input_per_mtok": 0.27, "output_per_mtok": 1.10}),
]
```
- **`estimate_cost(model: str, input_tokens: int, output_tokens: int) -> float`**: Calculate cost in USD
- **Returns 0.0 for unknown models** - safe default
- **`estimate_cost(model: str, input_tokens: int, output_tokens: int) -> float`**: Uses regex match, returns 0.0 for unknown models
#### Token Tracking in ai_client.py
- **`_add_bleed_derived()`** (ai_client.py): Adds derived token counts to comms entries
- **`get_history_bleed_stats()`**: Returns token statistics from history
- **Gemini**: Token counts from API response (`usage_metadata`)
- **Anthropic**: Token counts from API response (`usage`)
- **DeepSeek**: Token counts from API response (`usage`)
#### MMA Tier Usage Tracking
- **`ConductorEngine.tier_usage`** (multi_agent_conductor.py): Tracks per-tier token usage
#### MMA Tier Usage Tracking (multi_agent_conductor.py)
- **`ConductorEngine.tier_usage`** already tracks per-tier token counts AND model:
```python
self.tier_usage = {
"Tier 1": {"input": 0, "output": 0},
"Tier 2": {"input": 0, "output": 0},
"Tier 3": {"input": 0, "output": 0},
"Tier 4": {"input": 0, "output": 0},
"Tier 1": {"input": 0, "output": 0, "model": "gemini-3.1-pro-preview"},
"Tier 2": {"input": 0, "output": 0, "model": "gemini-3-flash-preview"},
"Tier 3": {"input": 0, "output": 0, "model": "gemini-2.5-flash-lite"},
"Tier 4": {"input": 0, "output": 0, "model": "gemini-2.5-flash-lite"},
}
```
- **Key insight**: The model name is already tracked per tier in `tier_usage[tier]["model"]`
- Updated in `run_worker_lifecycle()` from comms_log token counts
### Gaps to Fill (This Track's Scope)
- No GUI panel to display cost information
- No session-level cost accumulation
- No per-model breakdown visualization
- No tier breakdown visualization
- No per-tier cost breakdown in UI
## Architectural Constraints
### Non-Blocking Updates
- Cost calculations MUST NOT block UI thread
- Token counts are read from existing tracking - no new API calls
- Token counts are read from existing tier_usage - no new tracking needed
- Use cached values, update on state change events
### Cross-Thread Data Access
- `tier_usage` is updated on asyncio worker thread
- GUI reads via `_process_pending_gui_tasks` pattern
- Already synchronized through MMA state updates
### Memory Efficiency
- Session cost is a simple float - no history array needed
- Per-model costs can be dict: `{model_name: float}`
- `tier_usage` is updated on worker threads
- GUI reads via MMA state updates through `_pending_gui_tasks` pattern
- Already synchronized through existing state update mechanism
## Architecture Reference
@@ -69,38 +58,72 @@ Real-time cost tracking panel displaying cost per model, session totals, and bre
| File | Lines | Purpose |
|------|-------|---------|
| `src/cost_tracker.py` | 10-40 | `MODEL_PRICING`, `estimate_cost()` |
| `src/ai_client.py` | ~500-550 | `_add_bleed_derived()`, `get_history_bleed_stats()` |
| `src/multi_agent_conductor.py` | ~50-60 | `tier_usage` dict |
| `src/multi_agent_conductor.py` | ~50-60 | `tier_usage` dict with input/output/model |
| `src/gui_2.py` | ~2700-2800 | `_render_mma_dashboard()` - existing tier usage display |
| `src/gui_2.py` | ~1800-1900 | `_render_token_budget_panel()` - potential location |
### Existing MMA Dashboard Pattern
The `_render_mma_dashboard()` method already displays tier usage in a table. Extend this pattern for cost display.
### Cost Calculation Pattern
```python
from src import cost_tracker
usage = engine.tier_usage["Tier 3"]
cost = cost_tracker.estimate_cost(
usage["model"], # Already tracked!
usage["input"],
usage["output"]
)
```
## Functional Requirements
### FR1: Session Cost Accumulation
- Track total cost for the current session
- Track total cost for the current session in App/AppController state
- Reset on session reset
- Store in `App` or `AppController` state
- Sum of all tier costs
### FR2: Per-Model Cost Display
- Show cost broken down by model name
- Group by provider (Gemini, Anthropic, DeepSeek)
- Show token counts alongside costs
### FR2: Per-Tier Cost Display
- Show cost per MMA tier using existing `tier_usage[tier]["model"]` for model
- Show input/output tokens alongside cost
- Calculate using `cost_tracker.estimate_cost()`
### FR3: Tier Breakdown Display
- Show cost per MMA tier (Tier 1-4)
- Use existing `tier_usage` data
- Calculate cost using `cost_tracker.estimate_cost()`
### FR4: Real-Time Updates
### FR3: Real-Time Updates
- Update cost display when MMA state changes
- Hook into existing `mma_state_update` event handling
- No polling - event-driven
## Non-Functional Requirements
| Requirement | Constraint |
|-------------|------------|
| Frame Time Impact | <1ms when panel visible |
| Memory Overhead | <1KB for session cost state |
## Testing Requirements
### Unit Tests
- Test `estimate_cost()` with known model/token combinations
- Test unknown model returns 0.0
- Test session cost accumulation
### Integration Tests (via `live_gui` fixture)
- Verify cost panel displays after MMA execution
- Verify session reset clears costs
## Out of Scope
- Historical cost tracking across sessions
- Cost budgeting/alerts
- Per-model aggregation (model already per-tier)
## Acceptance Criteria
- [ ] Cost panel displays in GUI
- [ ] Per-tier cost shown with token counts
- [ ] Tier breakdown uses existing tier_usage model field
- [ ] Total session cost accumulates correctly
- [ ] Panel updates on MMA state changes
- [ ] Uses existing `cost_tracker.estimate_cost()`
- [ ] Session reset clears costs
- [ ] 1-space indentation maintained
## Non-Functional Requirements
| Requirement | Constraint |
|-------------|------------|
| Frame Time Impact | <1ms when panel visible |

130
conductor/tracks/kill_abort_workers_20260306/spec.md
View File

@@ -1,49 +1,40 @@
# Track Specification: Kill/Abort Running Workers (kill_abort_workers_20260306)
## Overview
Add ability to kill/abort a running Tier 3 worker mid-execution. Currently workers run to completion in `run_in_executor()`; add cancel button with forced termination option.
Add ability to kill/abort a running Tier 3 worker mid-execution. Currently workers run to completion; add cancel button with forced termination option.
## Current State Audit
### Already Implemented (DO NOT re-implement)
#### Worker Execution (multi_agent_conductor.py)
- **`run_worker_lifecycle()`**: Executes ticket via `loop.run_in_executor(None, ...)`
- **`ConductorEngine.run()`**: Main loop that spawns workers
- **No cancellation mechanism exists**
- **`run_worker_lifecycle()`**: Executes ticket via `threading.Thread(daemon=True)`
- **`ConductorEngine.run()`**: Spawns parallel workers:
- **No thread references stored** - threads launched and joined() but no tracking
- **No abort mechanism** - no way to stop a running worker
#### Thread Pool Pattern (app_controller.py / gui_2.py)
- Uses `threading.Thread(daemon=True)` for background work
- `asyncio.run_in_executor()` for blocking AI calls
- No thread references stored for later cancellation
#### Threading (multi_agent_conductor.py)
- **`threading.Thread`**: Used for workers
- **`threading.Event`**: Available for signaling
- **No abort event per worker**
#### GUI State (gui_2.py)
- **`mma_streams`**: Dict of worker output streams by tier
- **`active_tickets`**: List of currently active tickets
- **No per-worker thread tracking**
### Gaps to Fill (This Track's Scope)
- No way to track individual worker threads
- No cancellation signal mechanism
- No UI for kill/abort
### Gaps to Fill (This Track's scope)
- No worker thread tracking
- No abort signal mechanism
- No kill button UI
- No cleanup on termination
## Architectural Constraints
### Thread Safety
- Worker tracking MUST use thread-safe data structures
- Kill signal MUST be atomic (threading.Event)
- Status updates MUST be atomic
### Clean Termination
- Resources (file handles, network connections) MUST be released
- Partial results SHOULD be preserved
- No zombie processes
### AI Client Cancellation
- `ai_client.send()` is blocking - cannot be interrupted mid-call
- Kill can only happen between API calls or during tool execution
- Use `threading.Event` to signal abort between operations
### Abort Timing
- **AI API calls cannot mid-call interruption** (API limitation)
- Abort only between API calls or during tool execution
- Check abort flag between operations
## Architecture Reference
@@ -51,56 +42,85 @@ Add ability to kill/abort a running Tier 3 worker mid-execution. Currently worke
| File | Lines | Purpose |
|------|-------|---------|
| `src/multi_agent_conductor.py` | 150-250 | `run_worker_lifecycle()` - add abort check |
| `src/multi_agent_conductor.py` | 80-120 | `ConductorEngine.run()` - track workers |
| `src/dag_engine.py` | 50-80 | `ExecutionEngine` - add kill method |
| `src/models.py` | 30-50 | `Ticket` - add abort_event field |
| `src/gui_2.py` | 2700-2800 | `_render_mma_dashboard()` - add kill buttons |
### Proposed Worker Tracking Pattern
| `src/multi_agent_conductor.py` | ~80-150 | `ConductorEngine.run()` - thread spawning |
| `src/multi_agent_conductor.py` | ~250-320 | `run_worker_lifecycle()` - add abort check |
| `src/gui_2.py` | ~2650-2750 | `_render_mma_dashboard()` - add kill buttons |
### Current Thread Pattern
```python
# In ConductorEngine:
self._active_workers: dict[str, dict] = {} # ticket_id -> {thread, event, status}
# In ConductorEngine.run():
threads = []
for ticket in to_run:
t = threading.Thread(
target=run_worker_lifecycle,
args=(ticket, context, context_files, self.event_queue, self, md_content),
daemon=True
)
threads.append(t)
t.start()
# In Ticket:
self._abort_event: threading.Event = threading.Event()
# In run_worker_lifecycle:
if ticket._abort_event.is_set():
return # Exit early
for t in threads:
t.join()
```
## Functional Requirements
### FR1: Worker Thread Tracking
- Store reference to each worker's thread
- Store abort Event per worker
- Track status: running, killing, killed
- Store thread reference in `_active_workers: dict[ticket_id, Thread]`
- Track thread state: running, completed, killed
- Clean up on completion
### FR2: Kill Button UI
- Button per active worker in MMA dashboard
### FR2: Abort Event Mechan
- Add `threading.Event()` per ticket: `_abort_events[ticket_id]`
- Worker checks event between operations:
- API call cannot be interrupted (limitation documented)
### FR3: Kill Button UI
- Button per running worker in MMA dashboard
- Confirmation dialog before kill
- Disabled if no workers running
### FR3: Abort Signal Mechanism
- `threading.Event` per ticket for abort signaling
- Worker checks event between operations
- AI client call cannot be interrupted (document limitation)
### FR4: Clean Cleanup
- Mark ticket as "killed" status
### FR4: Clean Termination
- On kill: set `abort_event.set()`
- Wait for thread to finish (with timeout)
- Remove from `_active_workers`
- Preserve partial output in stream
- Remove from active workers dict
## Non-Functional Requirements
| Requirement | Constraint |
|-------------|------------|
| Response Time | Kill takes effect within 1s of button press |
| No Deadlocks | Kill cannot cause system hang |
| Memory Safety | Worker resources freed after kill |
## Testing Requirements
### Unit Tests
- Test abort event stops worker at check point
- Test worker tracking dict updates correctly
- Test kill button enables/disables based on workers
### Integration Tests (via `live_gui` fixture)
- Start worker, click kill, verify termination
- Verify partial output preserved
- Verify no zombie threads
## Out of Scope
- Force-killing AI API calls (API limitation)
- Kill and restart (separate track)
- Kill during PowerShell execution (separate concern)
## Acceptance Criteria
- [ ] Kill button visible per running worker
- [ ] Confirmation dialog appears
- [ ] Worker terminates within 1s of kill
- [ ] Partial output preserved in stream
- [ ] Resources cleaned up
- [ ] Status reflects "killed"
- [ ] No zombie threads after kill
- [ ] 1-space indentation maintained
| No Deadlocks | Kill cannot cause system hang |
| Memory Safety | Worker resources freed after kill |
## Testing Requirements
### Unit Tests