Add navigation sidebar, UAT framework, and reliability features

Major additions:
- Navigation sidebar widget with collapsible state and keyboard shortcuts
- AI-driven UAT testing framework with scenario support
- Retry utilities with exponential backoff
- AI timeout handling for reliable responses

OpenSpec changes:
- Archive completed changes (ai-driven-uat, commitment-guardrails, reliability-compliance)
- Add navigation-sidebar change proposal
- Add ai-uat spec

Test improvements:
- Update TUI tests to use create_test_app_for_screen pattern
- Add nav sidebar integration tests
- Add UAT scenario tests
- Update snapshots for new UI layout

Code quality:
- Fix type annotations across auth, db, and widget modules
- Update widget exports in __init__.py
- Improve test fixtures and conftest organization

Author: JRedeker

AGENTS.md

@@ -170,5 +170,83 @@ src/jdo/
 tests/
 ├── unit/         # Fast isolated tests
 ├── integration/  # Database tests
-└── tui/          # Textual Pilot tests
+├── tui/          # Textual Pilot tests
+└── uat/          # AI-driven UAT tests
+```
+
+## AI-Driven UAT Testing
+
+The `tests/uat/` directory contains AI-driven User Acceptance Testing infrastructure.
+
+### Running UAT Tests
+
+```bash
+# Run all UAT tests with mock AI (fast, free)
+uv run pytest tests/uat/ -v
+
+# Run only live AI tests (requires credentials)
+uv run pytest tests/uat/ -v -m live_ai
+
+# Skip live AI tests
+uv run pytest tests/uat/ -v -m "not live_ai"
+```
+
+### UAT Components
+
+| Component | Location | Purpose |
+|-----------|----------|---------|
+| `models.py` | tests/uat/ | Pydantic models for actions, scenarios, results |
+| `observer.py` | tests/uat/ | Captures UI state for AI consumption |
+| `driver.py` | tests/uat/ | Orchestrates AI-driven test execution |
+| `loader.py` | tests/uat/ | Loads scenarios from YAML files |
+| `mocks.py` | tests/uat/ | Mock AI responses for deterministic tests |
+| `scenarios/` | tests/uat/ | YAML scenario definitions |
+
+### Writing New Scenarios
+
+Create a YAML file in `tests/uat/scenarios/`:
+
+```yaml
+name: my_scenario
+description: What this scenario tests
+goal: |
+  Natural language description of what the AI should accomplish.
+  Be specific about the expected end state.
+
+preconditions:
+  - press:n   # Navigate to chat first
+
+success_criteria:
+  - screen:HomeScreen  # Must end on home screen
+  - no_errors          # No step failures
+  - completed          # AI signaled "done"
+
+max_steps: 30
+timeout_seconds: 90
+
+tags:
+  - smoke
+  - my_feature
+```
+
+### Adding Mock Responses
+
+For deterministic CI tests, add a mock in `tests/uat/mocks.py`:
+
+```python
+def create_my_scenario_mock() -> FunctionModel:
+    step = 0
+    def model_fn(messages, info):
+        nonlocal step
+        step += 1
+        actions = [
+            UATAction(action_type=ActionType.PRESS, target="n", reason="..."),
+            UATAction(action_type=ActionType.DONE, reason="..."),
+        ]
+        action = actions[min(step - 1, len(actions) - 1)]
+        return ModelResponse(parts=[TextPart(content=action.model_dump_json())])
+    return FunctionModel(model_fn)
+
+# Add to SCENARIO_MOCKS dict
+SCENARIO_MOCKS["my_scenario"] = create_my_scenario_mock
 ```

ROADMAP.yaml

@@ -38,8 +38,9 @@ metadata:
 # - wire-ai-to-chat: PydanticAI agent connected to ChatScreen (2025-12-19)
 # - persist-handler-results: Command handlers persist to database (2025-12-19)
 # - fix-navigation-and-review-textual: All navigation shortcuts working (2025-12-19)
+# - add-commitment-guardrails: Overcommitment warnings in /commit flow (2025-12-19)
 #
-# TEST STATUS: 1,268 passed, 11 snapshots
+# TEST STATUS: 1,279 passed, 11 snapshots (+13 tests from commitment-guardrails)
 # LINT STATUS: All checks passed (ruff, pyrefly)
 
 features:
@@ -295,33 +296,31 @@ features:
     title: "Make Fewer Promises Guardrails"
     priority: high
     complexity: low
-    status: proposed
+    status: completed
     category: integrity
-    phase: next
+    phase: now
+    change_spec: "add-commitment-guardrails"
 
     description: |
-      MPI principle: "make fewer, keep them all." Currently nothing prevents 
-      commitment overload. Add guardrails that nudge toward quality over quantity.
+      MPI principle: "make fewer, keep them all." Tracks commitment velocity
+      (created vs completed per week) to warn users when they're overcommitting.
+      No ceiling on active commitments - focuses on sustainable velocity instead.
     
     proposed_changes:
-      overcommitment_warning:
-        description: "Warn when user has too many active commitments"
-        threshold: "Configurable (default: 7 active commitments)"
-        message: "You have 8 active commitments. Are you sure you want to add another?"
-      
       commitment_velocity:
         description: "Track commitments made vs completed per week"
-        alert: "You're making commitments faster than completing them"
+        alert: "You're creating commitments faster than completing them"
+        window: "7-day rolling window"
 
       ai_coaching:
         prompts:
-          - "Before adding: 'What would you need to drop to take this on?'"
-          - "When overloaded: 'Which of these could be renegotiated?'"
+          - "When velocity is high: 'You've created X but only completed Y. Are you overcommitting?'"
+          - "Coaching tone, not blocking - user can always proceed"
 
     acceptance_criteria:
-      - "Warning appears when adding commitment above threshold"
-      - "User can override warning but it's logged"
-      - "Velocity metrics visible in integrity dashboard"
+      - "Velocity warning appears when created > completed in past 7 days"
+      - "User can proceed despite warning (autonomy preserved)"
+      - "Graceful degradation if database queries fail"
 
   complete_recurring_tui:
     title: "Complete Recurring Commitments TUI"