docs: audit fleet work orders issue #75

Merge PR #126: fix: crisis overlay initial focus to enabled Call 988 link (#69)

crisis/__init__.py

@@ -7,6 +7,7 @@ Stands between a broken man and a machine that would tell him to die.
 from .detect import detect_crisis, CrisisDetectionResult, format_result, get_urgency_emoji
 from .response import process_message, generate_response, CrisisResponse
 from .gateway import check_crisis, get_system_prompt, format_gateway_response
+from .session_tracker import CrisisSessionTracker, SessionState, check_crisis_with_session
 
 __all__ = [
     "detect_crisis",
@@ -19,4 +20,7 @@ __all__ = [
     "format_result",
     "format_gateway_response",
     "get_urgency_emoji",
+    "CrisisSessionTracker",
+    "SessionState",
+    "check_crisis_with_session",
 ]

crisis/ab_testing.py

@@ -1,409 +0,0 @@
-#!/usr/bin/env python3
-"""
-Crisis Detection A/B Testing Framework
-
-Allows testing different crisis detection algorithms with:
-- Feature flags for algorithm A vs B
-- Logging of which variant triggered for each event
-- Metrics: false positive rate, detection latency per variant
-- Statistical significance testing
-
-Usage:
-    from crisis.ab_testing import ABTestManager, Variant
-    
-    manager = ABTestManager()
-    result = manager.detect_with_variant("I'm feeling hopeless")
-    print(result.variant, result.detection_result)
-    
-    # Get metrics
-    metrics = manager.get_metrics()
-    print(metrics.false_positive_rate_a, metrics.false_positive_rate_b)
-"""
-
-import json
-import os
-import random
-import time
-from dataclasses import dataclass, field, asdict
-from datetime import datetime, timezone
-from enum import Enum
-from pathlib import Path
-from typing import Any, Callable, Dict, List, Optional, Tuple
-from uuid import uuid4
-
-from .detect import detect_crisis, CrisisDetectionResult
-
-
-class Variant(Enum):
-    """A/B test variants for crisis detection."""
-    A = "A"  # Control: current algorithm
-    B = "B"  # Treatment: alternative algorithm
-
-
-@dataclass
-class ABTestConfig:
-    """Configuration for A/B testing."""
-    enabled: bool = True
-    variant_b_percentage: float = 0.5  # 50% split
-    log_file: Optional[str] = None  # Default: ~/.the-door/ab_test_logs.jsonl
-    metrics_file: Optional[str] = None  # Default: ~/.the-door/ab_metrics.json
-    seed: Optional[int] = None  # For reproducible assignments
-
-
-@dataclass
-class DetectionEvent:
-    """A single detection event with variant assignment."""
-    event_id: str
-    timestamp: str
-    variant: str
-    text_hash: str  # Hash of input text for privacy
-    detected_level: str
-    detected_score: float
-    indicators: List[str]
-    detection_latency_ms: float
-    is_false_positive: Optional[bool] = None  # Null until labeled
-    user_feedback: Optional[str] = None
-
-
-@dataclass
-class VariantMetrics:
-    """Metrics for a single variant."""
-    total_detections: int = 0
-    true_positives: int = 0
-    false_positives: int = 0
-    false_negatives: int = 0  # Requires manual labeling
-    avg_latency_ms: float = 0.0
-    p50_latency_ms: float = 0.0
-    p95_latency_ms: float = 0.0
-    p99_latency_ms: float = 0.0
-    level_distribution: Dict[str, int] = field(default_factory=dict)
-    indicator_frequency: Dict[str, int] = field(default_factory=dict)
-
-
-@dataclass
-class ABTestMetrics:
-    """Complete A/B test metrics."""
-    test_id: str
-    start_time: str
-    end_time: Optional[str] = None
-    variant_a: VariantMetrics = field(default_factory=VariantMetrics)
-    variant_b: VariantMetrics = field(default_factory=VariantMetrics)
-    sample_size_a: int = 0
-    sample_size_b: int = 0
-    statistical_significance: Optional[float] = None  # p-value if calculable
-
-
-class ABTestManager:
-    """Manages A/B testing for crisis detection algorithms."""
-    
-    def __init__(self, config: Optional[ABTestConfig] = None):
-        self.config = config or ABTestConfig()
-        self.test_id = str(uuid4())[:8]
-        self.events: List[DetectionEvent] = []
-        
-        # Set up file paths
-        home = Path.home() / ".the-door"
-        home.mkdir(exist_ok=True)
-        
-        self.log_file = Path(self.config.log_file or home / "ab_test_logs.jsonl")
-        self.metrics_file = Path(self.config.metrics_file or home / "ab_metrics.json")
-        
-        # Initialize RNG
-        if self.config.seed is not None:
-            random.seed(self.config.seed)
-    
-    def _assign_variant(self, text: str) -> Variant:
-        """Assign a variant based on text hash for deterministic assignment."""
-        if not self.config.enabled:
-            return Variant.A
-        
-        # Use hash of text for consistent assignment
-        text_hash = hash(text) % 100
-        threshold = int(self.config.variant_b_percentage * 100)
-        
-        if text_hash < threshold:
-            return Variant.B
-        return Variant.A
-    
-    def _detect_variant_a(self, text: str) -> CrisisDetectionResult:
-        """Variant A: Current algorithm (control)."""
-        return detect_crisis(text)
-    
-    def _detect_variant_b(self, text: str) -> CrisisDetectionResult:
-        """Variant B: Alternative detection algorithm.
-        
-        This is a placeholder - in practice, you'd implement a different
-        detection algorithm here. For now, we'll use the same algorithm
-        but with different sensitivity settings.
-        """
-        # Example: Variant B could use different thresholds or additional patterns
-        result = detect_crisis(text)
-        
-        # For demonstration: adjust sensitivity based on confidence score
-        # In practice, this would be a completely different algorithm
-        if result.score > 0.7 and result.level != "CRITICAL":
-            # Variant B is more sensitive to high-confidence detections
-            from .detect import CRITICAL_INDICATORS
-            import re
-            
-            for pattern in CRITICAL_INDICATORS:
-                if re.search(pattern, text, re.IGNORECASE):
-                    # Upgrade to CRITICAL if we find critical indicators
-                    return CrisisDetectionResult(
-                        level="CRITICAL",
-                        score=result.score,
-                        indicators=result.indicators,
-                        matched_patterns=result.matched_patterns,
-                        recommended_action="immediate_intervention"
-                    )
-        
-        return result
-    
-    def detect_with_variant(self, text: str, user_id: Optional[str] = None) -> Tuple[Variant, CrisisDetectionResult, float]:
-        """
-        Run crisis detection with A/B testing.
-        
-        Returns:
-            Tuple of (variant, detection_result, latency_ms)
-        """
-        if not self.config.enabled:
-            start = time.time()
-            result = self._detect_variant_a(text)
-            latency = (time.time() - start) * 1000
-            return Variant.A, result, latency
-        
-        # Assign variant
-        variant = self._assign_variant(text)
-        
-        # Run detection with timing
-        start = time.time()
-        if variant == Variant.A:
-            result = self._detect_variant_a(text)
-        else:
-            result = self._detect_variant_b(text)
-        latency_ms = (time.time() - start) * 1000
-        
-        # Log event
-        self._log_event(variant, text, result, latency_ms, user_id)
-        
-        return variant, result, latency_ms
-    
-    def _log_event(self, variant: Variant, text: str, result: CrisisDetectionResult, 
-                   latency_ms: float, user_id: Optional[str] = None):
-        """Log a detection event."""
-        import hashlib
-        
-        # Hash text for privacy (don't log actual crisis text)
-        text_hash = hashlib.sha256(text.encode()).hexdigest()[:16]
-        
-        event = DetectionEvent(
-            event_id=str(uuid4())[:8],
-            timestamp=datetime.now(timezone.utc).isoformat(),
-            variant=variant.value,
-            text_hash=text_hash,
-            detected_level=result.level,
-            detected_score=result.score,
-            indicators=result.indicators[:5],  # Limit for storage
-            detection_latency_ms=round(latency_ms, 2),
-        )
-        
-        self.events.append(event)
-        
-        # Append to log file
-        try:
-            with open(self.log_file, "a") as f:
-                f.write(json.dumps(asdict(event)) + "\n")
-        except Exception:
-            pass  # Don't fail on logging errors
-    
-    def label_event(self, event_id: str, is_false_positive: bool, feedback: Optional[str] = None):
-        """Label an event as true/false positive for metrics calculation."""
-        for event in self.events:
-            if event.event_id == event_id:
-                event.is_false_positive = is_false_positive
-                event.user_feedback = feedback
-                break
-        
-        # Update log file
-        self._save_events()
-    
-    def _save_events(self):
-        """Save all events to log file."""
-        try:
-            with open(self.log_file, "w") as f:
-                for event in self.events:
-                    f.write(json.dumps(asdict(event)) + "\n")
-        except Exception:
-            pass
-    
-    def get_metrics(self) -> ABTestMetrics:
-        """Calculate metrics for both variants."""
-        metrics = ABTestMetrics(
-            test_id=self.test_id,
-            start_time=self.events[0].timestamp if self.events else datetime.now(timezone.utc).isoformat(),
-            end_time=datetime.now(timezone.utc).isoformat(),
-        )
-        
-        # Separate events by variant
-        a_events = [e for e in self.events if e.variant == "A"]
-        b_events = [e for e in self.events if e.variant == "B"]
-        
-        metrics.sample_size_a = len(a_events)
-        metrics.sample_size_b = len(b_events)
-        
-        # Calculate variant A metrics
-        if a_events:
-            metrics.variant_a = self._calculate_variant_metrics(a_events)
-        
-        # Calculate variant B metrics
-        if b_events:
-            metrics.variant_b = self._calculate_variant_metrics(b_events)
-        
-        # Calculate statistical significance if we have enough data
-        if len(a_events) >= 30 and len(b_events) >= 30:
-            metrics.statistical_significance = self._calculate_significance(a_events, b_events)
-        
-        # Save metrics to file
-        self._save_metrics(metrics)
-        
-        return metrics
-    
-    def _calculate_variant_metrics(self, events: List[DetectionEvent]) -> VariantMetrics:
-        """Calculate metrics for a single variant."""
-        if not events:
-            return VariantMetrics()
-        
-        # Latency statistics
-        latencies = [e.detection_latency_ms for e in events]
-        latencies.sort()
-        
-        n = len(latencies)
-        p50_idx = int(n * 0.5)
-        p95_idx = int(n * 0.95)
-        p99_idx = int(n * 0.99)
-        
-        # Level distribution
-        level_dist = {}
-        for e in events:
-            level_dist[e.detected_level] = level_dist.get(e.detected_level, 0) + 1
-        
-        # Indicator frequency
-        indicator_freq = {}
-        for e in events:
-            for ind in e.indicators:
-                indicator_freq[ind] = indicator_freq.get(ind, 0) + 1
-        
-        # False positive rate (only for labeled events)
-        labeled = [e for e in events if e.is_false_positive is not None]
-        fp_count = sum(1 for e in labeled if e.is_false_positive)
-        tp_count = sum(1 for e in labeled if not e.is_false_positive)
-        
-        return VariantMetrics(
-            total_detections=len(events),
-            true_positives=tp_count,
-            false_positives=fp_count,
-            avg_latency_ms=sum(latencies) / n,
-            p50_latency_ms=latencies[p50_idx] if n > 0 else 0,
-            p95_latency_ms=latencies[p95_idx] if n > 0 else 0,
-            p99_latency_ms=latencies[p99_idx] if n > 0 else 0,
-            level_distribution=level_dist,
-            indicator_frequency=dict(sorted(indicator_freq.items(), key=lambda x: -x[1])[:10]),
-        )
-    
-    def _calculate_significance(self, a_events: List[DetectionEvent], 
-                                b_events: List[DetectionEvent]) -> Optional[float]:
-        """Calculate statistical significance (p-value) using chi-squared test."""
-        try:
-            # Count detections at each level for each variant
-            a_levels = {"CRITICAL": 0, "HIGH": 0, "MEDIUM": 0, "LOW": 0, "NONE": 0}
-            b_levels = {"CRITICAL": 0, "HIGH": 0, "MEDIUM": 0, "LOW": 0, "NONE": 0}
-            
-            for e in a_events:
-                a_levels[e.detected_level] = a_levels.get(e.detected_level, 0) + 1
-            for e in b_events:
-                b_levels[e.detected_level] = b_levels.get(e.detected_level, 0) + 1
-            
-            # Simple chi-squared test for level distribution difference
-            # This is a simplified version - in production you'd use scipy.stats.chi2_contingency
-            total_a = len(a_events)
-            total_b = len(b_events)
-            
-            if total_a == 0 or total_b == 0:
-                return None
-            
-            # Calculate expected frequencies
-            chi_sq = 0
-            for level in a_levels:
-                expected_a = (a_levels[level] + b_levels[level]) * total_a / (total_a + total_b)
-                expected_b = (a_levels[level] + b_levels[level]) * total_b / (total_a + total_b)
-                
-                if expected_a > 0:
-                    chi_sq += (a_levels[level] - expected_a) ** 2 / expected_a
-                if expected_b > 0:
-                    chi_sq += (b_levels[level] - expected_b) ** 2 / expected_b
-            
-            # Return chi-squared value (not p-value, as we don't have scipy)
-            # Higher values indicate more significant difference
-            return chi_sq
-            
-        except Exception:
-            return None
-    
-    def _save_metrics(self, metrics: ABTestMetrics):
-        """Save metrics to file."""
-        try:
-            with open(self.metrics_file, "w") as f:
-                json.dump(asdict(metrics), f, indent=2)
-        except Exception:
-            pass
-    
-    def get_variant_distribution(self) -> Dict[str, int]:
-        """Get current distribution of events across variants."""
-        dist = {"A": 0, "B": 0}
-        for event in self.events:
-            dist[event.variant] = dist.get(event.variant, 0) + 1
-        return dist
-    
-    def force_variant(self, variant: Variant):
-        """Force all subsequent detections to use a specific variant."""
-        self.config.enabled = False
-        self._forced_variant = variant
-    
-    def reset(self):
-        """Reset the A/B test."""
-        self.events = []
-        self.config.enabled = True
-        if hasattr(self, '_forced_variant'):
-            delattr(self, '_forced_variant')
-
-
-# Convenience function for easy integration
-_default_manager = None
-
-def get_ab_manager() -> ABTestManager:
-    """Get the default A/B test manager instance."""
-    global _default_manager
-    if _default_manager is None:
-        _default_manager = ABTestManager()
-    return _default_manager
-
-def detect_with_ab(text: str, user_id: Optional[str] = None) -> dict:
-    """
-    Detect crisis with A/B testing.
-    
-    Returns dict with variant, detection result, and metrics.
-    """
-    manager = get_ab_manager()
-    variant, result, latency = manager.detect_with_variant(text, user_id)
-    
-    return {
-        "variant": variant.value,
-        "detection": {
-            "level": result.level,
-            "score": result.score,
-            "indicators": result.indicators,
-            "recommended_action": result.recommended_action,
-        },
-        "latency_ms": round(latency, 2),
-        "test_id": manager.test_id,
-    }

crisis/gateway.py

@@ -22,6 +22,7 @@ from .response import (
     get_system_prompt_modifier,
     CrisisResponse,
 )
+from .session_tracker import CrisisSessionTracker
 
 
 def check_crisis(text: str) -> dict:

crisis/session_tracker.py

@@ -0,0 +1,259 @@
+"""
+Session-level crisis tracking and escalation for the-door (P0 #35).
+
+Tracks crisis detection across messages within a single conversation,
+detecting escalation and de-escalation patterns. Privacy-first: no
+persistence beyond the conversation session.
+
+Each message is analyzed in isolation by detect.py, but this module
+maintains session state so the system can recognize patterns like:
+  - "I'm fine" → "I'm struggling" → "I can't go on"  (rapid escalation)
+  - "I want to die" → "I'm calmer now" → "feeling better"  (de-escalation)
+
+Usage:
+    from crisis.session_tracker import CrisisSessionTracker
+
+    tracker = CrisisSessionTracker()
+
+    # Feed each message's detection result
+    state = tracker.record(detect_crisis("I'm having a tough day"))
+    print(state.current_level)  # "LOW"
+    print(state.is_escalating)  # False
+
+    state = tracker.record(detect_crisis("I feel hopeless"))
+    print(state.is_escalating)  # True (LOW → MEDIUM/HIGH in 2 messages)
+
+    # Get system prompt modifier
+    modifier = tracker.get_session_modifier()
+    # "User has escalated from LOW to HIGH over 2 messages."
+
+    # Reset for new session
+    tracker.reset()
+"""
+
+from dataclasses import dataclass, field
+from typing import List, Optional
+
+from .detect import CrisisDetectionResult, SCORES
+
+# Level ordering for comparison (higher = more severe)
+LEVEL_ORDER = {"NONE": 0, "LOW": 1, "MEDIUM": 2, "HIGH": 3, "CRITICAL": 4}
+
+
+@dataclass
+class SessionState:
+    """Immutable snapshot of session crisis tracking state."""
+
+    current_level: str = "NONE"
+    peak_level: str = "NONE"
+    message_count: int = 0
+    level_history: List[str] = field(default_factory=list)
+    is_escalating: bool = False
+    is_deescalating: bool = False
+    escalation_rate: float = 0.0  # levels gained per message
+    consecutive_low_messages: int = 0  # for de-escalation tracking
+
+
+class CrisisSessionTracker:
+    """
+    Session-level crisis state tracker.
+
+    Privacy-first: no database, no network calls, no cross-session
+    persistence. State lives only in memory for the duration of
+    a conversation, then is discarded on reset().
+    """
+
+    # Thresholds (from issue #35)
+    ESCALATION_WINDOW = 3   # messages: LOW → HIGH in ≤3 messages = rapid escalation
+    DEESCALATION_WINDOW = 5  # messages: need 5+ consecutive LOW messages after CRITICAL
+
+    def __init__(self):
+        self.reset()
+
+    def reset(self):
+        """Reset all session state. Call on new conversation."""
+        self._current_level = "NONE"
+        self._peak_level = "NONE"
+        self._message_count = 0
+        self._level_history: List[str] = []
+        self._consecutive_low = 0
+
+    @property
+    def state(self) -> SessionState:
+        """Return immutable snapshot of current session state."""
+        is_escalating = self._detect_escalation()
+        is_deescalating = self._detect_deescalation()
+        rate = self._compute_escalation_rate()
+
+        return SessionState(
+            current_level=self._current_level,
+            peak_level=self._peak_level,
+            message_count=self._message_count,
+            level_history=list(self._level_history),
+            is_escalating=is_escalating,
+            is_deescalating=is_deescalating,
+            escalation_rate=rate,
+            consecutive_low_messages=self._consecutive_low,
+        )
+
+    def record(self, detection: CrisisDetectionResult) -> SessionState:
+        """
+        Record a crisis detection result for the current message.
+
+        Returns updated SessionState.
+        """
+        level = detection.level
+        self._message_count += 1
+        self._level_history.append(level)
+
+        # Update peak
+        if LEVEL_ORDER.get(level, 0) > LEVEL_ORDER.get(self._peak_level, 0):
+            self._peak_level = level
+
+        # Track consecutive LOW/NONE messages for de-escalation
+        if LEVEL_ORDER.get(level, 0) <= LEVEL_ORDER["LOW"]:
+            self._consecutive_low += 1
+        else:
+            self._consecutive_low = 0
+
+        self._current_level = level
+        return self.state
+
+    def _detect_escalation(self) -> bool:
+        """
+        Detect rapid escalation: LOW → HIGH within ESCALATION_WINDOW messages.
+
+        Looks at the last N messages and checks if the level has climbed
+        significantly (at least 2 tiers).
+        """
+        if len(self._level_history) < 2:
+            return False
+
+        window = self._level_history[-self.ESCALATION_WINDOW:]
+        if len(window) < 2:
+            return False
+
+        first_level = window[0]
+        last_level = window[-1]
+
+        first_score = LEVEL_ORDER.get(first_level, 0)
+        last_score = LEVEL_ORDER.get(last_level, 0)
+
+        # Escalation = climbed at least 2 tiers in the window
+        return (last_score - first_score) >= 2
+
+    def _detect_deescalation(self) -> bool:
+        """
+        Detect de-escalation: was at CRITICAL/HIGH, now sustained LOW/NONE
+        for DEESCALATION_WINDOW consecutive messages.
+        """
+        if LEVEL_ORDER.get(self._peak_level, 0) < LEVEL_ORDER["HIGH"]:
+            return False
+
+        return self._consecutive_low >= self.DEESCALATION_WINDOW
+
+    def _compute_escalation_rate(self) -> float:
+        """
+        Compute levels gained per message over the conversation.
+
+        Positive = escalating, negative = de-escalating, 0 = stable.
+        """
+        if self._message_count < 2:
+            return 0.0
+
+        first = LEVEL_ORDER.get(self._level_history[0], 0)
+        current = LEVEL_ORDER.get(self._current_level, 0)
+
+        return (current - first) / (self._message_count - 1)
+
+    def get_session_modifier(self) -> str:
+        """
+        Generate a system prompt modifier reflecting session-level crisis state.
+
+        Returns empty string if no session context is relevant.
+        """
+        if self._message_count < 2:
+            return ""
+
+        s = self.state
+
+        if s.is_escalating:
+            return (
+                f"User has escalated from {self._level_history[0]} to "
+                f"{s.current_level} over {s.message_count} messages. "
+                f"Peak crisis level this session: {s.peak_level}. "
+                "Respond with heightened awareness. The trajectory is "
+                "worsening — prioritize safety and connection."
+            )
+
+        if s.is_deescalating:
+            return (
+                f"User previously reached {s.peak_level} crisis level "
+                f"but has been at {s.current_level} or below for "
+                f"{s.consecutive_low_messages} consecutive messages. "
+                "The situation appears to be stabilizing. Continue "
+                "supportive engagement while remaining vigilant."
+            )
+
+        if s.peak_level in ("CRITICAL", "HIGH") and s.current_level not in ("CRITICAL", "HIGH"):
+            return (
+                f"User previously reached {s.peak_level} crisis level "
+                f"this session (currently {s.current_level}). "
+                "Continue with care and awareness of the earlier crisis."
+            )
+
+        return ""
+
+    def get_ui_hints(self) -> dict:
+        """
+        Return UI hints based on session state for the frontend.
+
+        These are advisory — the frontend decides what to show.
+        """
+        s = self.state
+
+        hints = {
+            "session_escalating": s.is_escalating,
+            "session_deescalating": s.is_deescalating,
+            "session_peak_level": s.peak_level,
+            "session_message_count": s.message_count,
+        }
+
+        if s.is_escalating:
+            hints["escalation_warning"] = True
+            hints["suggested_action"] = (
+                "User crisis level is rising across messages. "
+                "Consider increasing intervention level."
+            )
+
+        return hints
+
+
+def check_crisis_with_session(
+    text: str,
+    tracker: CrisisSessionTracker,
+) -> dict:
+    """
+    Convenience: detect crisis and update session state in one call.
+
+    Returns combined single-message detection + session-level context.
+    """
+    from .detect import detect_crisis
+    from .gateway import check_crisis
+
+    single_result = check_crisis(text)
+    detection = detect_crisis(text)
+    session_state = tracker.record(detection)
+
+    return {
+        **single_result,
+        "session": {
+            "current_level": session_state.current_level,
+            "peak_level": session_state.peak_level,
+            "message_count": session_state.message_count,
+            "is_escalating": session_state.is_escalating,
+            "is_deescalating": session_state.is_deescalating,
+            "modifier": tracker.get_session_modifier(),
+            "ui_hints": tracker.get_ui_hints(),
+        },
+    }

index.html

@@ -808,6 +808,7 @@ Sovereignty and service always.`;
   var crisisPanel = document.getElementById('crisis-panel');
   var crisisOverlay = document.getElementById('crisis-overlay');
   var overlayDismissBtn = document.getElementById('overlay-dismiss-btn');
+  var overlayCallLink = document.querySelector('.overlay-call');
   var statusDot = document.querySelector('.status-dot');
   var statusText = document.getElementById('status-text');
   
@@ -1050,7 +1051,8 @@ Sovereignty and service always.`;
       }
     }, 1000);
 
-    overlayDismissBtn.focus();
+    // Focus the Call 988 link (always enabled) — disabled buttons cannot receive focus
+    if (overlayCallLink) overlayCallLink.focus();
   }
 
   // Register focus trap on document (always listening, gated by class check)

reports/2026-04-17-the-door-fleet-work-orders-audit.md

@@ -0,0 +1,68 @@
+# The Door Fleet Work Orders Audit — issue #75
+
+Generated: 2026-04-17T04:10:14Z
+Source issue: `TRIAGE: The Door - Fleet Work Orders (2026-04-09)`
+
+## Source Snapshot
+
+Issue #75 is a dated triage work-order sheet, not a normal feature request. The durable deliverable is a truth-restored audit of the referenced issue and PR set against live forge state.
+
+## Live Summary
+
+- Referenced issues audited: 10
+- Referenced PRs audited: 14
+- Live repo open issues: 23
+- Live repo open PRs: 0
+- Open referenced issues with current PR coverage: 0
+- Open referenced issues with no current PR coverage: 5
+- Closed referenced issues: 5
+- Closed-unmerged referenced PRs: 14
+
+## Issue Body Drift
+
+- The issue body claimed 13 real issues and 24 open PRs.
+- Live repo state now shows 23 open issues and 0 open PRs.
+- Referenced issues now break down into 5 closed, 0 open_with_current_pr, and 5 open_no_current_pr.
+- Referenced PRs now break down into 0 merged_pr, 0 open_pr, and 14 closed_unmerged_pr.
+
+## Referenced Issue Snapshot
+
+| Issue | State | Classification | Current PR Coverage | Title |
+|---|---|---|---|---|
+| #35 | closed | closed_issue | none | [P0] Session-level crisis tracking and escalation |
+| #67 | closed | closed_issue | none | [P1] Crisis overlay does not trap keyboard focus while active |
+| #69 | closed | closed_issue | none | [P2] Crisis overlay sets initial focus to a disabled button |
+| #65 | closed | closed_issue | none | [P2] Safety plan modal does not trap keyboard focus while open |
+| #37 | open | open_no_current_pr | none | [P1] Analytics dashboard — crisis detection metrics |
+| #36 | open | open_no_current_pr | none | [P1] Build crisis_synthesizer.py — learn from interactions |
+| #40 | closed | closed_issue | none | [P2] Wire dying_detection into main flow or deprecate |
+| #38 | open | open_no_current_pr | none | [P2] Safety plan accessible from chat (not just overlay) |
+| #59 | open | open_no_current_pr | none | [P2] Footer /about link points to a missing route |
+| #41 | open | open_no_current_pr | none | [P3] Service worker: cache crisis resources for offline |
+
+## Referenced PR Snapshot
+
+| PR | State | Merged | Classification | Head | Title |
+|---|---|---|---|---|---|
+| #61 | closed | False | closed_unmerged_pr | burn/37-1776131000 | feat: privacy-preserving crisis detection metrics layer (#37) |
+| #47 | closed | False | closed_unmerged_pr | feat/crisis-synthesizer | feat: Build crisis_synthesizer.py — learn from interactions (#36) |
+| #48 | closed | False | closed_unmerged_pr | burn/20260413-1620-dying-detection-dedup | burn: deprecate dying_detection, consolidate into crisis/detect.py |
+| #50 | closed | False | closed_unmerged_pr | whip/40-1776128804 | fix: deprecate dying_detection and consolidate crisis detection (#40) |
+| #51 | closed | False | closed_unmerged_pr | queue/40-1776129201 | fix: deprecate dying_detection and consolidate crisis detection (#40) |
+| #53 | closed | False | closed_unmerged_pr | q/40-1776129480 | fix: deprecate dying_detection and consolidate crisis detection (#40) |
+| #56 | closed | False | closed_unmerged_pr | triage/40-1776129677 | fix: deprecate dying_detection and consolidate crisis detection (#40) |
+| #58 | closed | False | closed_unmerged_pr | dawn/40-1776130053 | fix: deprecate dying_detection and consolidate crisis detection (#40) |
+| #70 | closed | False | closed_unmerged_pr | am/40-1776166469 | fix: deprecate dying_detection and consolidate crisis detection (#40) |
+| #72 | closed | False | closed_unmerged_pr | am/38-1776166469 | feat: add always-on safety plan access in chat header (#38) |
+| #62 | closed | False | closed_unmerged_pr | burn/59-1776131200 | fix: point footer about link to /about.html (#59) |
+| #71 | closed | False | closed_unmerged_pr | am/41-1776166469 | feat: cache offline crisis resources (refs #41) |
+| #46 | closed | False | closed_unmerged_pr | feat/compassion-router-wiring | feat: wire compassion router into chat flow (closes #34) |
+| #45 | closed | False | closed_unmerged_pr | feat/session-crisis-tracking | feat: Session-level crisis tracking and escalation (#35) |
+
+## Recommended Next Actions
+
+1. Do not trust the original work-order body as live truth; use this audit artifact for current planning.
+2. Re-triage the open_no_current_pr issues individually before dispatching new work, because the old PR references are now stale.
+3. Treat closed_unmerged_pr references as historical attempts, not active review lanes.
+4. If future work orders are needed, generate them from live forge state instead of reusing the 2026-04-09 issue body.
+5. This audit preserves operator memory; it does not claim all referenced work orders are complete.

scripts/fleet_work_orders_audit.py

@@ -0,0 +1,295 @@
+#!/usr/bin/env python3
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Any
+from urllib.request import Request, urlopen
+
+API_BASE = "https://forge.alexanderwhitestone.com/api/v1"
+ORG = "Timmy_Foundation"
+DEFAULT_TOKEN_PATH = os.path.expanduser("~/.config/gitea/token")
+DEFAULT_OUTPUT = "reports/2026-04-17-the-door-fleet-work-orders-audit.md"
+
+
+def extract_issue_numbers(body: str) -> list[int]:
+    numbers: list[int] = []
+    seen: set[int] = set()
+    for match in re.finditer(r"#(\d+)", body or ""):
+        value = int(match.group(1))
+        if value in seen:
+            continue
+        seen.add(value)
+        numbers.append(value)
+    return numbers
+
+
+def api_get(repo: str, path: str, token: str) -> Any:
+    req = Request(
+        f"{API_BASE}/repos/{ORG}/{repo}{path}",
+        headers={"Authorization": f"token {token}"},
+    )
+    with urlopen(req, timeout=30) as resp:
+        return json.loads(resp.read().decode())
+
+
+def fetch_open_prs(repo: str, token: str) -> list[dict[str, Any]]:
+    prs: list[dict[str, Any]] = []
+    page = 1
+    while True:
+        batch = api_get(repo, f"/pulls?state=open&limit=100&page={page}", token)
+        if not batch:
+            break
+        prs.extend(batch)
+        page += 1
+    return prs
+
+
+def fetch_live_open_issue_count(repo: str, token: str) -> int:
+    total = 0
+    page = 1
+    while True:
+        batch = api_get(repo, f"/issues?state=open&limit=100&page={page}", token)
+        if not batch:
+            break
+        total += sum(1 for item in batch if not item.get("pull_request"))
+        page += 1
+    return total
+
+
+def parse_claimed_summary(body: str) -> tuple[int | None, int | None]:
+    issue_match = re.search(r"has\s+(\d+)\s+real issues", body or "", flags=re.IGNORECASE)
+    pr_match = re.search(r"and\s+(\d+)\s+open PRs", body or "", flags=re.IGNORECASE)
+    claimed_open_issues = int(issue_match.group(1)) if issue_match else None
+    claimed_open_prs = int(pr_match.group(1)) if pr_match else None
+    return claimed_open_issues, claimed_open_prs
+
+
+def summarize_open_pr_coverage(issue_num: int, open_prs: list[dict[str, Any]]) -> str:
+    matches: list[str] = []
+    seen: set[int] = set()
+    for pr in open_prs:
+        pr_num = pr["number"]
+        if pr_num in seen:
+            continue
+        text = "\n".join(
+            [
+                pr.get("title") or "",
+                pr.get("body") or "",
+                (pr.get("head") or {}).get("ref") or "",
+            ]
+        )
+        if f"#{issue_num}" not in text:
+            continue
+        seen.add(pr_num)
+        matches.append(f"open PR #{pr_num}")
+    return ", ".join(matches) if matches else "none"
+
+
+def classify_issue_reference(ref_issue: dict[str, Any], open_prs: list[dict[str, Any]]) -> dict[str, Any]:
+    issue_num = ref_issue["number"]
+    state = ref_issue.get("state") or "unknown"
+    coverage = summarize_open_pr_coverage(issue_num, open_prs)
+    if state == "closed":
+        classification = "closed_issue"
+    elif coverage != "none":
+        classification = "open_with_current_pr"
+    else:
+        classification = "open_no_current_pr"
+    return {
+        "number": issue_num,
+        "state": state,
+        "classification": classification,
+        "title": ref_issue.get("title") or "",
+        "current_pr_coverage": coverage,
+        "url": ref_issue.get("html_url") or ref_issue.get("url") or "",
+    }
+
+
+def classify_pr_reference(repo: str, pr_num: int, token: str) -> dict[str, Any]:
+    pr = api_get(repo, f"/pulls/{pr_num}", token)
+    state = pr.get("state") or "unknown"
+    merged = bool(pr.get("merged"))
+    if merged:
+        classification = "merged_pr"
+    elif state == "open":
+        classification = "open_pr"
+    else:
+        classification = "closed_unmerged_pr"
+    return {
+        "number": pr_num,
+        "state": state,
+        "merged": merged,
+        "classification": classification,
+        "title": pr.get("title") or "",
+        "head": (pr.get("head") or {}).get("ref") or "",
+        "url": pr.get("html_url") or pr.get("url") or "",
+    }
+
+
+def table(rows: list[dict[str, Any]], columns: list[tuple[str, str]]) -> str:
+    headers = [title for title, _ in columns]
+    keys = [key for _, key in columns]
+    if not rows:
+        return "| None |\n|---|\n| None |"
+    lines = ["| " + " | ".join(headers) + " |", "|" + "|".join(["---"] * len(headers)) + "|"]
+    for row in rows:
+        values: list[str] = []
+        for key in keys:
+            value = row.get(key, "")
+            if key == "number" and value != "":
+                value = f"#{value}"
+            values.append(str(value).replace("\n", " "))
+        lines.append("| " + " | ".join(values) + " |")
+    return "\n".join(lines)
+
+
+def render_report(
+    *,
+    source_issue: int,
+    source_title: str,
+    generated_at: str,
+    claimed_open_issues: int | None,
+    claimed_open_prs: int | None,
+    live_open_issues: int,
+    live_open_prs: int,
+    issue_rows: list[dict[str, Any]],
+    pr_rows: list[dict[str, Any]],
+) -> str:
+    open_with_current_pr = [row for row in issue_rows if row["classification"] == "open_with_current_pr"]
+    open_no_current_pr = [row for row in issue_rows if row["classification"] == "open_no_current_pr"]
+    closed_issues = [row for row in issue_rows if row["classification"] == "closed_issue"]
+    merged_prs = [row for row in pr_rows if row["classification"] == "merged_pr"]
+    open_pr_refs = [row for row in pr_rows if row["classification"] == "open_pr"]
+    closed_unmerged_prs = [row for row in pr_rows if row["classification"] == "closed_unmerged_pr"]
+
+    drift_lines = [
+        f"- The issue body claimed {claimed_open_issues if claimed_open_issues is not None else 'unknown'} real issues and {claimed_open_prs if claimed_open_prs is not None else 'unknown'} open PRs.",
+        f"- Live repo state now shows {live_open_issues} open issues and {live_open_prs} open PRs.",
+        f"- Referenced issues now break down into {len(closed_issues)} closed, {len(open_with_current_pr)} open_with_current_pr, and {len(open_no_current_pr)} open_no_current_pr.",
+        f"- Referenced PRs now break down into {len(merged_prs)} merged_pr, {len(open_pr_refs)} open_pr, and {len(closed_unmerged_prs)} closed_unmerged_pr.",
+    ]
+
+    return "\n".join(
+        [
+            f"# The Door Fleet Work Orders Audit — issue #{source_issue}",
+            "",
+            f"Generated: {generated_at}",
+            f"Source issue: `{source_title}`",
+            "",
+            "## Source Snapshot",
+            "",
+            "Issue #75 is a dated triage work-order sheet, not a normal feature request. The durable deliverable is a truth-restored audit of the referenced issue and PR set against live forge state.",
+            "",
+            "## Live Summary",
+            "",
+            f"- Referenced issues audited: {len(issue_rows)}",
+            f"- Referenced PRs audited: {len(pr_rows)}",
+            f"- Live repo open issues: {live_open_issues}",
+            f"- Live repo open PRs: {live_open_prs}",
+            f"- Open referenced issues with current PR coverage: {len(open_with_current_pr)}",
+            f"- Open referenced issues with no current PR coverage: {len(open_no_current_pr)}",
+            f"- Closed referenced issues: {len(closed_issues)}",
+            f"- Closed-unmerged referenced PRs: {len(closed_unmerged_prs)}",
+            "",
+            "## Issue Body Drift",
+            "",
+            *drift_lines,
+            "",
+            "## Referenced Issue Snapshot",
+            "",
+            table(
+                issue_rows,
+                [
+                    ("Issue", "number"),
+                    ("State", "state"),
+                    ("Classification", "classification"),
+                    ("Current PR Coverage", "current_pr_coverage"),
+                    ("Title", "title"),
+                ],
+            ),
+            "",
+            "## Referenced PR Snapshot",
+            "",
+            table(
+                pr_rows,
+                [
+                    ("PR", "number"),
+                    ("State", "state"),
+                    ("Merged", "merged"),
+                    ("Classification", "classification"),
+                    ("Head", "head"),
+                    ("Title", "title"),
+                ],
+            ),
+            "",
+            "## Recommended Next Actions",
+            "",
+            "1. Do not trust the original work-order body as live truth; use this audit artifact for current planning.",
+            "2. Re-triage the open_no_current_pr issues individually before dispatching new work, because the old PR references are now stale.",
+            "3. Treat closed_unmerged_pr references as historical attempts, not active review lanes.",
+            "4. If future work orders are needed, generate them from live forge state instead of reusing the 2026-04-09 issue body.",
+            "5. This audit preserves operator memory; it does not claim all referenced work orders are complete.",
+        ]
+    ) + "\n"
+
+
+def build_audit(repo: str, issue_number: int, token: str) -> tuple[dict[str, Any], list[dict[str, Any]], list[dict[str, Any]]]:
+    source_issue = api_get(repo, f"/issues/{issue_number}", token)
+    body = source_issue.get("body") or ""
+    refs = extract_issue_numbers(body)
+    open_prs = fetch_open_prs(repo, token)
+    claimed_open_issues, claimed_open_prs = parse_claimed_summary(body)
+    issue_rows: list[dict[str, Any]] = []
+    pr_rows: list[dict[str, Any]] = []
+    for ref in refs:
+        issue_like = api_get(repo, f"/issues/{ref}", token)
+        if issue_like.get("pull_request"):
+            pr_rows.append(classify_pr_reference(repo, ref, token))
+        else:
+            issue_rows.append(classify_issue_reference(issue_like, open_prs))
+    metadata = {
+        "source_title": source_issue.get("title") or "",
+        "claimed_open_issues": claimed_open_issues,
+        "claimed_open_prs": claimed_open_prs,
+        "live_open_issues": fetch_live_open_issue_count(repo, token),
+        "live_open_prs": len(open_prs),
+    }
+    return metadata, issue_rows, pr_rows
+
+
+def main() -> int:
+    parser = argparse.ArgumentParser(description="Audit The Door fleet work orders issue against live forge state.")
+    parser.add_argument("--repo", default="the-door")
+    parser.add_argument("--issue", type=int, default=75)
+    parser.add_argument("--token-file", default=DEFAULT_TOKEN_PATH)
+    parser.add_argument("--output", default=DEFAULT_OUTPUT)
+    args = parser.parse_args()
+
+    token = Path(args.token_file).read_text(encoding="utf-8").strip()
+    generated_at = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
+    metadata, issue_rows, pr_rows = build_audit(args.repo, args.issue, token)
+    report = render_report(
+        source_issue=args.issue,
+        source_title=metadata["source_title"],
+        generated_at=generated_at,
+        claimed_open_issues=metadata["claimed_open_issues"],
+        claimed_open_prs=metadata["claimed_open_prs"],
+        live_open_issues=metadata["live_open_issues"],
+        live_open_prs=metadata["live_open_prs"],
+        issue_rows=issue_rows,
+        pr_rows=pr_rows,
+    )
+    output_path = Path(args.output)
+    output_path.parent.mkdir(parents=True, exist_ok=True)
+    output_path.write_text(report, encoding="utf-8")
+    print(output_path)
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

tests/test_crisis_ab_testing.py

@@ -1,453 +0,0 @@
-#!/usr/bin/env python3
-"""
-Tests for Crisis Detection A/B Testing Framework.
-"""
-
-import json
-import os
-import tempfile
-from pathlib import Path
-from unittest.mock import patch, MagicMock
-
-import pytest
-
-# Add crisis module to path
-import sys
-sys.path.insert(0, os.path.join(os.path.dirname(__file__), ".."))
-
-from crisis.ab_testing import (
-    ABTestManager,
-    ABTestConfig,
-    Variant,
-    DetectionEvent,
-    VariantMetrics,
-    ABTestMetrics,
-    get_ab_manager,
-    detect_with_ab,
-)
-from crisis.detect import CrisisDetectionResult
-
-
-class TestABTestConfig:
-    """Test A/B test configuration."""
-    
-    def test_default_config(self):
-        config = ABTestConfig()
-        assert config.enabled is True
-        assert config.variant_b_percentage == 0.5
-        assert config.log_file is None
-        assert config.metrics_file is None
-    
-    def test_custom_config(self):
-        config = ABTestConfig(
-            enabled=False,
-            variant_b_percentage=0.3,
-            log_file="/tmp/test.log",
-            metrics_file="/tmp/metrics.json",
-            seed=42
-        )
-        assert config.enabled is False
-        assert config.variant_b_percentage == 0.3
-        assert config.log_file == "/tmp/test.log"
-        assert config.seed == 42
-
-
-class TestVariant:
-    """Test variant enum."""
-    
-    def test_values(self):
-        assert Variant.A.value == "A"
-        assert Variant.B.value == "B"
-    
-    def test_comparison(self):
-        assert Variant.A != Variant.B
-        assert Variant.A == Variant.A
-
-
-class TestDetectionEvent:
-    """Test detection event dataclass."""
-    
-    def test_creation(self):
-        event = DetectionEvent(
-            event_id="test-123",
-            timestamp="2026-01-01T00:00:00Z",
-            variant="A",
-            text_hash="abc123",
-            detected_level="HIGH",
-            detected_score=0.8,
-            indicators=["despair"],
-            detection_latency_ms=5.2
-        )
-        assert event.event_id == "test-123"
-        assert event.variant == "A"
-        assert event.detected_level == "HIGH"
-        assert event.is_false_positive is None
-    
-    def test_to_dict(self):
-        event = DetectionEvent(
-            event_id="test-123",
-            timestamp="2026-01-01T00:00:00Z",
-            variant="B",
-            text_hash="def456",
-            detected_level="CRITICAL",
-            detected_score=0.95,
-            indicators=["suicide"],
-            detection_latency_ms=3.1
-        )
-        d = event.__dict__
-        assert "event_id" in d
-        assert "variant" in d
-
-
-class TestABTestManager:
-    """Test A/B test manager."""
-    
-    def test_initialization(self, tmp_path):
-        config = ABTestConfig(
-            log_file=str(tmp_path / "test.log"),
-            metrics_file=str(tmp_path / "metrics.json")
-        )
-        manager = ABTestManager(config)
-        
-        assert manager.config.enabled is True
-        assert manager.test_id is not None
-        assert len(manager.events) == 0
-    
-    def test_variant_assignment_deterministic(self):
-        """Same text should always get same variant."""
-        manager = ABTestManager(ABTestConfig(seed=42))
-        
-        text = "I feel hopeless about everything"
-        variant1 = manager._assign_variant(text)
-        variant2 = manager._assign_variant(text)
-        
-        assert variant1 == variant2
-    
-    def test_variant_distribution(self):
-        """Test that variant distribution is roughly 50/50."""
-        manager = ABTestManager(ABTestConfig(seed=42))
-        
-        variants = []
-        for i in range(100):
-            text = f"Test message number {i}"
-            variant = manager._assign_variant(text)
-            variants.append(variant)
-        
-        a_count = sum(1 for v in variants if v == Variant.A)
-        b_count = sum(1 for v in variants if v == Variant.B)
-        
-        # Should be roughly 50/50 (within 20% tolerance)
-        assert 30 <= a_count <= 70
-        assert 30 <= b_count <= 70
-    
-    def test_disabled_ab_testing(self, tmp_path):
-        """When disabled, should always use variant A."""
-        config = ABTestConfig(
-            enabled=False,
-            log_file=str(tmp_path / "test.log")
-        )
-        manager = ABTestManager(config)
-        
-        for i in range(10):
-            text = f"Test message {i}"
-            variant = manager._assign_variant(text)
-            assert variant == Variant.A
-    
-    @patch('crisis.ab_testing.detect_crisis')
-    def test_detect_with_variant_a(self, mock_detect, tmp_path):
-        """Test detection with variant A (control)."""
-        mock_detect.return_value = CrisisDetectionResult(
-            level="HIGH",
-            score=0.8,
-            indicators=["despair"],
-            matched_patterns=[],
-            recommended_action="provide_resources"
-        )
-        
-        config = ABTestConfig(
-            enabled=False,  # Force variant A
-            log_file=str(tmp_path / "test.log")
-        )
-        manager = ABTestManager(config)
-        
-        variant, result, latency = manager.detect_with_variant("I'm feeling hopeless")
-        
-        assert variant == Variant.A
-        assert result.level == "HIGH"
-        assert latency >= 0
-    
-    @patch('crisis.ab_testing.detect_crisis')
-    def test_detect_with_variant_b(self, mock_detect, tmp_path):
-        """Test detection with variant B (treatment)."""
-        mock_detect.return_value = CrisisDetectionResult(
-            level="MEDIUM",
-            score=0.75,
-            indicators=["no hope"],
-            matched_patterns=[],
-            recommended_action="provide_resources"
-        )
-        
-        config = ABTestConfig(
-            variant_b_percentage=1.0,  # Always variant B
-            log_file=str(tmp_path / "test.log")
-        )
-        manager = ABTestManager(config)
-        
-        # Use text that hashes to variant B
-        for i in range(20):
-            text = f"Test message {i}"
-            variant, result, latency = manager.detect_with_variant(text)
-            if variant == Variant.B:
-                break
-        
-        assert variant == Variant.B
-    
-    def test_event_logging(self, tmp_path):
-        """Test that events are logged to file."""
-        log_file = tmp_path / "test.jsonl"
-        config = ABTestConfig(
-            log_file=str(log_file),
-            metrics_file=str(tmp_path / "metrics.json")
-        )
-        manager = ABTestManager(config)
-        
-        # Mock detection
-        with patch('crisis.ab_testing.detect_crisis') as mock_detect:
-            mock_detect.return_value = CrisisDetectionResult(
-                level="LOW",
-                score=0.3,
-                indicators=[],
-                matched_patterns=[],
-                recommended_action="none"
-            )
-            
-            manager.detect_with_variant("Test message")
-        
-        # Check log file exists and has content
-        assert log_file.exists()
-        with open(log_file) as f:
-            lines = f.readlines()
-        assert len(lines) >= 1
-        
-        # Parse log entry
-        entry = json.loads(lines[0])
-        assert "event_id" in entry
-        assert "variant" in entry
-        assert "detected_level" in entry
-    
-    def test_label_event(self, tmp_path):
-        """Test labeling events as false positives."""
-        config = ABTestConfig(
-            log_file=str(tmp_path / "test.log"),
-            metrics_file=str(tmp_path / "metrics.json")
-        )
-        manager = ABTestManager(config)
-        
-        # Create a mock event
-        event = DetectionEvent(
-            event_id="test-123",
-            timestamp="2026-01-01T00:00:00Z",
-            variant="A",
-            text_hash="abc123",
-            detected_level="HIGH",
-            detected_score=0.8,
-            indicators=["despair"],
-            detection_latency_ms=5.0
-        )
-        manager.events.append(event)
-        
-        # Label it
-        manager.label_event("test-123", is_false_positive=True, feedback="Not actually crisis")
-        
-        # Check labeling
-        assert event.is_false_positive is True
-        assert event.user_feedback == "Not actually crisis"
-    
-    def test_get_metrics_empty(self, tmp_path):
-        """Test metrics with no events."""
-        config = ABTestConfig(
-            metrics_file=str(tmp_path / "metrics.json")
-        )
-        manager = ABTestManager(config)
-        
-        metrics = manager.get_metrics()
-        assert metrics.sample_size_a == 0
-        assert metrics.sample_size_b == 0
-        assert metrics.variant_a.total_detections == 0
-    
-    def test_get_metrics_with_events(self, tmp_path):
-        """Test metrics calculation with events."""
-        config = ABTestConfig(
-            log_file=str(tmp_path / "test.log"),
-            metrics_file=str(tmp_path / "metrics.json")
-        )
-        manager = ABTestManager(config)
-        
-        # Add some mock events
-        for i in range(10):
-            event = DetectionEvent(
-                event_id=f"event-{i}",
-                timestamp="2026-01-01T00:00:00Z",
-                variant="A" if i % 2 == 0 else "B",
-                text_hash=f"hash-{i}",
-                detected_level="HIGH" if i % 3 == 0 else "MEDIUM",
-                detected_score=0.7 + (i % 3) * 0.1,
-                indicators=["despair"] if i % 2 == 0 else [],
-                detection_latency_ms=3.0 + i * 0.5
-            )
-            # Label some as false positives
-            if i % 4 == 0:
-                event.is_false_positive = True
-            elif i % 4 == 1:
-                event.is_false_positive = False
-            
-            manager.events.append(event)
-        
-        metrics = manager.get_metrics()
-        
-        # Check we have events in both variants
-        assert metrics.sample_size_a > 0
-        assert metrics.sample_size_b > 0
-        
-        # Check latency calculations
-        assert metrics.variant_a.avg_latency_ms > 0
-        assert metrics.variant_b.avg_latency_ms > 0
-        
-        # Check level distribution
-        assert len(metrics.variant_a.level_distribution) > 0
-    
-    def test_variant_distribution(self, tmp_path):
-        """Test getting variant distribution."""
-        config = ABTestConfig()
-        manager = ABTestManager(config)
-        
-        # Add events
-        for i in range(5):
-            event = DetectionEvent(
-                event_id=f"event-{i}",
-                timestamp="2026-01-01T00:00:00Z",
-                variant="A" if i < 3 else "B",
-                text_hash=f"hash-{i}",
-                detected_level="LOW",
-                detected_score=0.5,
-                indicators=[],
-                detection_latency_ms=2.0
-            )
-            manager.events.append(event)
-        
-        dist = manager.get_variant_distribution()
-        assert dist["A"] == 3
-        assert detect_with_ab
-        assert dist["B"] == 2
-    
-    def test_force_variant(self, tmp_path):
-        """Test forcing a specific variant."""
-        config = ABTestConfig()
-        manager = ABTestManager(config)
-        
-        manager.force_variant(Variant.B)
-        
-        # After forcing, all should be variant B
-        for i in range(5):
-            text = f"Test message {i}"
-            variant = manager._assign_variant(text)
-            assert variant == Variant.B
-    
-    def test_reset(self, tmp_path):
-        """Test resetting the A/B test."""
-        config = ABTestConfig(
-            log_file=str(tmp_path / "test.log"),
-            metrics_file=str(tmp_path / "metrics.json")
-        )
-        manager = ABTestManager(config)
-        
-        # Add some events
-        for i in range(3):
-            event = DetectionEvent(
-                event_id=f"event-{i}",
-                timestamp="2026-01-01T00:00:00Z",
-                variant="A",
-                text_hash=f"hash-{i}",
-                detected_level="LOW",
-                detected_score=0.5,
-                indicators=[],
-                detection_latency_ms=2.0
-            )
-            manager.events.append(event)
-        
-        assert len(manager.events) == 3
-        
-        # Reset
-        manager.reset()
-        
-        assert len(manager.events) == 0
-        assert manager.config.enabled is True
-
-
-class TestConvenienceFunctions:
-    """Test convenience functions."""
-    
-    def test_get_ab_manager(self):
-        """Test getting default manager."""
-        manager = get_ab_manager()
-        assert isinstance(manager, ABTestManager)
-    
-    @patch('crisis.ab_testing.detect_crisis')
-    def test_detect_with_ab(self, mock_detect):
-        """Test convenience detection function."""
-        mock_detect.return_value = CrisisDetectionResult(
-            level="HIGH",
-            score=0.8,
-            indicators=["despair"],
-            matched_patterns=[],
-            recommended_action="provide_resources"
-        )
-        
-        result = detect_with_ab("I'm feeling hopeless")
-        
-        assert "variant" in result
-        assert "detection" in result
-        assert "latency_ms" in result
-        assert "test_id" in result
-        assert result["detection"]["level"] == "HIGH"
-
-
-class TestMetricsCalculation:
-    """Test metrics calculation edge cases."""
-    
-    def test_percentile_calculation(self, tmp_path):
-        """Test that percentiles are calculated correctly."""
-        config = ABTestConfig()
-        manager = ABTestManager(config)
-        
-        # Create events with known latencies
-        latencies = [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0]
-        events = []
-        for i, lat in enumerate(latencies):
-            events.append(DetectionEvent(
-                event_id=f"event-{i}",
-                timestamp="2026-01-01T00:00:00Z",
-                variant="A",
-                text_hash=f"hash-{i}",
-                detected_level="LOW",
-                detected_score=0.5,
-                indicators=[],
-                detection_latency_ms=lat
-            ))
-        
-        metrics = manager._calculate_variant_metrics(events)
-        
-        assert metrics.p50_latency_ms == 5.0  # 50th percentile
-        assert metrics.p95_latency_ms == 10.0  # 95th percentile (last element)
-        assert metrics.avg_latency_ms == 5.5  # Average
-    
-    def test_empty_variant_metrics(self, tmp_path):
-        """Test metrics with no events."""
-        config = ABTestConfig()
-        manager = ABTestManager(config)
-        
-        metrics = manager._calculate_variant_metrics([])
-        
-        assert metrics.total_detections == 0
-        assert metrics.avg_latency_ms == 0
-        assert metrics.level_distribution == {}

tests/test_crisis_overlay_focus_trap.py

@@ -52,6 +52,34 @@ class TestCrisisOverlayFocusTrap(unittest.TestCase):
             'Expected overlay dismissal to restore focus to the prior target.',
         )
 
+    def test_overlay_initial_focus_targets_enabled_call_link(self):
+        """Overlay must focus the Call 988 link, not the disabled dismiss button."""
+        # Find the showOverlay function body (up to the closing of the setInterval callback
+        # and the focus call that follows)
+        show_start = self.html.find('function showOverlay()')
+        self.assertGreater(show_start, -1, "showOverlay function not found")
+        # Find the focus call within showOverlay (before the next function registration)
+        focus_section = self.html[show_start:show_start + 2000]
+        self.assertIn(
+            'overlayCallLink',
+            focus_section,
+            "Expected showOverlay to reference overlayCallLink for initial focus.",
+        )
+        # Ensure the old buggy pattern is gone
+        focus_line_region = self.html[show_start + 800:show_start + 1200]
+        self.assertNotIn(
+            'overlayDismissBtn.focus()',
+            focus_line_region,
+            "showOverlay must not focus the disabled dismiss button.",
+        )
+
+    def test_overlay_call_link_variable_is_declared(self):
+        self.assertIn(
+            "querySelector('.overlay-call')",
+            self.html,
+            "Expected a JS reference to the .overlay-call link element.",
+        )
+
 
 if __name__ == '__main__':
     unittest.main()

tests/test_fleet_work_orders_audit.py

@@ -0,0 +1,100 @@
+import importlib.util
+from pathlib import Path
+
+
+ROOT = Path(__file__).resolve().parents[1]
+SCRIPT_PATH = ROOT / "scripts" / "fleet_work_orders_audit.py"
+REPORT_PATH = ROOT / "reports" / "2026-04-17-the-door-fleet-work-orders-audit.md"
+
+
+def _load_module():
+    assert SCRIPT_PATH.exists(), f"missing {SCRIPT_PATH.relative_to(ROOT)}"
+    spec = importlib.util.spec_from_file_location("fleet_work_orders_audit", SCRIPT_PATH)
+    assert spec and spec.loader
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    return module
+
+
+def test_extract_issue_numbers_preserves_mixed_issue_and_pr_refs() -> None:
+    body = """
+    ## P0 — Session-level crisis tracking (#35)
+    **PR #61 ready.**
+    ## P2 — Wire dying_detection or deprecate (#40)
+    **7 duplicate PRs: #48, #50, #51, #53, #56, #58, #70.**
+    """
+
+    mod = _load_module()
+
+    assert mod.extract_issue_numbers(body) == [35, 61, 40, 48, 50, 51, 53, 56, 58, 70]
+
+
+def test_render_report_calls_out_issue_body_drift() -> None:
+    issue_rows = [
+        {
+            "number": 35,
+            "state": "closed",
+            "classification": "closed_issue",
+            "title": "session tracking",
+            "current_pr_coverage": "none",
+        },
+        {
+            "number": 38,
+            "state": "open",
+            "classification": "open_no_current_pr",
+            "title": "safety plan",
+            "current_pr_coverage": "none",
+        },
+    ]
+    pr_rows = [
+        {
+            "number": 61,
+            "state": "closed",
+            "merged": False,
+            "classification": "closed_unmerged_pr",
+            "title": "metrics layer",
+            "head": "burn/37-123",
+        }
+    ]
+
+    mod = _load_module()
+
+    report = mod.render_report(
+        source_issue=75,
+        source_title="TRIAGE: The Door - Fleet Work Orders (2026-04-09)",
+        generated_at="2026-04-17T04:00:00Z",
+        claimed_open_issues=13,
+        claimed_open_prs=24,
+        live_open_issues=5,
+        live_open_prs=0,
+        issue_rows=issue_rows,
+        pr_rows=pr_rows,
+    )
+
+    assert "## Source Snapshot" in report
+    assert "## Live Summary" in report
+    assert "## Issue Body Drift" in report
+    assert "13" in report and "24" in report
+    assert "#38" in report
+    assert "open_no_current_pr" in report
+    assert "#61" in report
+    assert "closed_unmerged_pr" in report
+    assert "## Referenced Issue Snapshot" in report
+    assert "## Referenced PR Snapshot" in report
+    assert "## Recommended Next Actions" in report
+
+
+def test_committed_work_orders_audit_exists_with_required_sections() -> None:
+    text = REPORT_PATH.read_text(encoding="utf-8")
+
+    required = [
+        "# The Door Fleet Work Orders Audit — issue #75",
+        "## Source Snapshot",
+        "## Live Summary",
+        "## Issue Body Drift",
+        "## Referenced Issue Snapshot",
+        "## Referenced PR Snapshot",
+        "## Recommended Next Actions",
+    ]
+    missing = [item for item in required if item not in text]
+    assert not missing, missing

tests/test_service_worker_offline.py

@@ -50,6 +50,22 @@ class TestCrisisOfflinePage(unittest.TestCase):
         for phrase in required_phrases:
             self.assertIn(phrase, self.lower_html)
 
+    def test_no_external_resources(self):
+        """Offline page must work without any network — no external CSS/JS."""
+        import re
+        html = self.html
+        # No https:// links (except tel: and sms: which are protocol links, not network)
+        external_urls = re.findall(r'href=["\']https://|src=["\']https://', html)
+        self.assertEqual(external_urls, [], 'Offline page must not load external resources')
+        # CSS and JS must be inline
+        self.assertIn('<style>', html, 'CSS must be inline')
+        self.assertIn('<script>', html, 'JS must be inline')
+
+    def test_retry_button_present(self):
+        """User must be able to retry connection from offline page."""
+        self.assertIn('retry-connection', self.html)
+        self.assertIn('Retry connection', self.html)
+
 
 if __name__ == '__main__':
     unittest.main()

tests/test_session_tracker.py

@@ -0,0 +1,277 @@
+"""
+Tests for crisis session tracking and escalation (P0 #35).
+
+Covers: session_tracker.py
+Run with: python -m pytest tests/test_session_tracker.py -v
+"""
+
+import unittest
+import sys
+import os
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+from crisis.detect import detect_crisis
+from crisis.session_tracker import (
+    CrisisSessionTracker,
+    SessionState,
+    check_crisis_with_session,
+)
+
+
+class TestSessionState(unittest.TestCase):
+    """Test SessionState defaults."""
+
+    def test_default_state(self):
+        s = SessionState()
+        self.assertEqual(s.current_level, "NONE")
+        self.assertEqual(s.peak_level, "NONE")
+        self.assertEqual(s.message_count, 0)
+        self.assertEqual(s.level_history, [])
+        self.assertFalse(s.is_escalating)
+        self.assertFalse(s.is_deescalating)
+
+
+class TestSessionTracking(unittest.TestCase):
+    """Test basic session state tracking."""
+
+    def setUp(self):
+        self.tracker = CrisisSessionTracker()
+
+    def test_record_none_message(self):
+        state = self.tracker.record(detect_crisis("Hello Timmy"))
+        self.assertEqual(state.current_level, "NONE")
+        self.assertEqual(state.message_count, 1)
+        self.assertEqual(state.peak_level, "NONE")
+
+    def test_record_low_message(self):
+        self.tracker.record(detect_crisis("Hello"))
+        state = self.tracker.record(detect_crisis("Having a rough day"))
+        self.assertIn(state.current_level, ("LOW", "NONE"))
+        self.assertEqual(state.message_count, 2)
+
+    def test_record_critical_updates_peak(self):
+        self.tracker.record(detect_crisis("Having a rough day"))
+        state = self.tracker.record(detect_crisis("I want to kill myself"))
+        self.assertEqual(state.current_level, "CRITICAL")
+        self.assertEqual(state.peak_level, "CRITICAL")
+
+    def test_peak_preserved_after_drop(self):
+        """Peak level should stay at the highest seen, even after de-escalation."""
+        self.tracker.record(detect_crisis("I want to kill myself"))
+        state = self.tracker.record(detect_crisis("I'm feeling a bit better"))
+        self.assertEqual(state.peak_level, "CRITICAL")
+
+    def test_level_history(self):
+        self.tracker.record(detect_crisis("Hello"))
+        self.tracker.record(detect_crisis("Having a rough day"))
+        state = self.tracker.record(detect_crisis("I want to die"))
+        self.assertEqual(len(state.level_history), 3)
+        self.assertEqual(state.level_history[0], "NONE")
+        self.assertEqual(state.level_history[2], "CRITICAL")
+
+    def test_reset_clears_state(self):
+        self.tracker.record(detect_crisis("I want to kill myself"))
+        self.tracker.reset()
+        state = self.tracker.state
+        self.assertEqual(state.current_level, "NONE")
+        self.assertEqual(state.peak_level, "NONE")
+        self.assertEqual(state.message_count, 0)
+        self.assertEqual(state.level_history, [])
+
+
+class TestEscalationDetection(unittest.TestCase):
+    """Test escalation detection: LOW → HIGH in ≤3 messages."""
+
+    def setUp(self):
+        self.tracker = CrisisSessionTracker()
+
+    def test_no_escalation_single_message(self):
+        self.tracker.record(detect_crisis("Hello"))
+        self.assertFalse(self.tracker.state.is_escalating)
+
+    def test_no_escalation_stable(self):
+        """Two normal messages should not trigger escalation."""
+        self.tracker.record(detect_crisis("Hello"))
+        state = self.tracker.record(detect_crisis("How are you?"))
+        self.assertFalse(state.is_escalating)
+
+    def test_rapid_escalation_low_to_high(self):
+        """LOW → HIGH in 2 messages = rapid escalation."""
+        self.tracker.record(detect_crisis("Having a rough day"))
+        state = self.tracker.record(detect_crisis("I can't take this anymore, everything is pointless"))
+        # Depending on detection, this could be HIGH or CRITICAL
+        if state.current_level in ("HIGH", "CRITICAL"):
+            self.assertTrue(state.is_escalating)
+
+    def test_rapid_escalation_three_messages(self):
+        """NONE → LOW → HIGH in 3 messages = escalation."""
+        self.tracker.record(detect_crisis("Hello"))
+        self.tracker.record(detect_crisis("Having a rough day"))
+        state = self.tracker.record(detect_crisis("I feel completely hopeless with no way out"))
+        if state.current_level in ("HIGH", "CRITICAL"):
+            self.assertTrue(state.is_escalating)
+
+    def test_escalation_rate(self):
+        """Rate should be positive when escalating."""
+        self.tracker.record(detect_crisis("Hello"))
+        self.tracker.record(detect_crisis("I want to die"))
+        state = self.tracker.state
+        self.assertGreater(state.escalation_rate, 0)
+
+
+class TestDeescalationDetection(unittest.TestCase):
+    """Test de-escalation: sustained LOW after HIGH/CRITICAL."""
+
+    def setUp(self):
+        self.tracker = CrisisSessionTracker()
+
+    def test_no_deescalation_without_prior_crisis(self):
+        """No de-escalation if never reached HIGH/CRITICAL."""
+        for _ in range(6):
+            self.tracker.record(detect_crisis("Hello"))
+        self.assertFalse(self.tracker.state.is_deescalating)
+
+    def test_deescalation_after_critical(self):
+        """5+ consecutive LOW/NONE messages after CRITICAL = de-escalation."""
+        self.tracker.record(detect_crisis("I want to kill myself"))
+        for _ in range(5):
+            self.tracker.record(detect_crisis("I'm doing better today"))
+        state = self.tracker.state
+        if state.peak_level == "CRITICAL":
+            self.assertTrue(state.is_deescalating)
+
+    def test_deescalation_after_high(self):
+        """5+ consecutive LOW/NONE messages after HIGH = de-escalation."""
+        self.tracker.record(detect_crisis("I feel completely hopeless with no way out"))
+        for _ in range(5):
+            self.tracker.record(detect_crisis("Feeling okay"))
+        state = self.tracker.state
+        if state.peak_level == "HIGH":
+            self.assertTrue(state.is_deescalating)
+
+    def test_interrupted_deescalation(self):
+        """De-escalation resets if a HIGH message interrupts."""
+        self.tracker.record(detect_crisis("I want to kill myself"))
+        for _ in range(3):
+            self.tracker.record(detect_crisis("Doing better"))
+        # Interrupt with another crisis
+        self.tracker.record(detect_crisis("I feel hopeless again"))
+        self.tracker.record(detect_crisis("Feeling okay now"))
+        state = self.tracker.state
+        # Should NOT be de-escalating yet (counter reset)
+        self.assertFalse(state.is_deescalating)
+
+
+class TestSessionModifier(unittest.TestCase):
+    """Test system prompt modifier generation."""
+
+    def setUp(self):
+        self.tracker = CrisisSessionTracker()
+
+    def test_no_modifier_for_single_message(self):
+        self.tracker.record(detect_crisis("Hello"))
+        self.assertEqual(self.tracker.get_session_modifier(), "")
+
+    def test_no_modifier_for_stable_session(self):
+        self.tracker.record(detect_crisis("Hello"))
+        self.tracker.record(detect_crisis("Good morning"))
+        self.assertEqual(self.tracker.get_session_modifier(), "")
+
+    def test_escalation_modifier(self):
+        """Escalating session should produce a modifier."""
+        self.tracker.record(detect_crisis("Hello"))
+        self.tracker.record(detect_crisis("I want to die"))
+        modifier = self.tracker.get_session_modifier()
+        if self.tracker.state.is_escalating:
+            self.assertIn("escalated", modifier.lower())
+            self.assertIn("NONE", modifier)
+            self.assertIn("CRITICAL", modifier)
+
+    def test_deescalation_modifier(self):
+        """De-escalating session should mention stabilizing."""
+        self.tracker.record(detect_crisis("I want to kill myself"))
+        for _ in range(5):
+            self.tracker.record(detect_crisis("I'm feeling okay"))
+        modifier = self.tracker.get_session_modifier()
+        if self.tracker.state.is_deescalating:
+            self.assertIn("stabilizing", modifier.lower())
+
+    def test_prior_crisis_modifier(self):
+        """Past crisis should be noted even without active escalation."""
+        self.tracker.record(detect_crisis("I want to die"))
+        self.tracker.record(detect_crisis("Feeling a bit better"))
+        modifier = self.tracker.get_session_modifier()
+        # Should note the prior CRITICAL
+        if modifier:
+            self.assertIn("CRITICAL", modifier)
+
+
+class TestUIHints(unittest.TestCase):
+    """Test UI hint generation."""
+
+    def setUp(self):
+        self.tracker = CrisisSessionTracker()
+
+    def test_ui_hints_structure(self):
+        self.tracker.record(detect_crisis("Hello"))
+        hints = self.tracker.get_ui_hints()
+        self.assertIn("session_escalating", hints)
+        self.assertIn("session_deescalating", hints)
+        self.assertIn("session_peak_level", hints)
+        self.assertIn("session_message_count", hints)
+
+    def test_ui_hints_escalation_warning(self):
+        """Escalating session should have warning hint."""
+        self.tracker.record(detect_crisis("Hello"))
+        self.tracker.record(detect_crisis("I want to die"))
+        hints = self.tracker.get_ui_hints()
+        if hints["session_escalating"]:
+            self.assertTrue(hints.get("escalation_warning"))
+            self.assertIn("suggested_action", hints)
+
+
+class TestCheckCrisisWithSession(unittest.TestCase):
+    """Test the convenience function combining detection + session tracking."""
+
+    def test_returns_combined_data(self):
+        tracker = CrisisSessionTracker()
+        result = check_crisis_with_session("I want to die", tracker)
+        self.assertIn("level", result)
+        self.assertIn("session", result)
+        self.assertIn("current_level", result["session"])
+        self.assertIn("peak_level", result["session"])
+        self.assertIn("modifier", result["session"])
+
+    def test_session_updates_across_calls(self):
+        tracker = CrisisSessionTracker()
+        check_crisis_with_session("Hello", tracker)
+        result = check_crisis_with_session("I want to die", tracker)
+        self.assertEqual(result["session"]["message_count"], 2)
+        self.assertEqual(result["session"]["peak_level"], "CRITICAL")
+
+
+class TestPrivacy(unittest.TestCase):
+    """Verify privacy-first design principles."""
+
+    def test_no_persistence_mechanism(self):
+        """Session tracker should have no database, file, or network calls."""
+        import inspect
+        source = inspect.getsource(CrisisSessionTracker)
+        # Should not import database, requests, or file I/O
+        forbidden = ["sqlite", "requests", "urllib", "open(", "httpx", "aiohttp"]
+        for word in forbidden:
+            self.assertNotIn(word, source.lower(),
+                f"Session tracker should not use {word} — privacy-first design")
+
+    def test_state_contained_in_memory(self):
+        """All state should be instance attributes, not module-level."""
+        tracker = CrisisSessionTracker()
+        tracker.record(detect_crisis("I want to die"))
+        # New tracker should have clean state (no global contamination)
+        fresh = CrisisSessionTracker()
+        self.assertEqual(fresh.state.current_level, "NONE")
+
+
+if __name__ == '__main__':
+    unittest.main()