Compare commits

..

1 Commits

Author SHA1 Message Date
Alexander Whitestone
6e03492147 feat: CLI command to view crisis metrics summary (#136)
crisis/metrics.py:
  CrisisMetrics class — aggregate crisis detection metrics
  Privacy-first: stores only counts, never user content
  Daily JSONL files in ~/.the-door/metrics/
  get_summary(days) → AggregateMetrics
  get_report(days) → human-readable report
  get_json(days) → JSON export
  CLI: python3 -m crisis.metrics --summary/--json

crisis/__init__.py:
  Export CrisisMetrics, AggregateMetrics

Makefile:
  make metrics      → summary report
  make metrics-json → JSON export

tests/test_crisis_metrics.py: 6 tests
  record_session, summary, report, JSON export
2026-04-17 01:26:44 -04:00
7 changed files with 380 additions and 53 deletions

View File

@@ -12,7 +12,7 @@ VPS := alexanderwhitestone.com
DOMAIN := alexanderwhitestone.com
DEPLOY_DIR := deploy
.PHONY: help deploy deploy-bash check ssl push service
.PHONY: help deploy deploy-bash check ssl push service metrics
help:
@echo "The Door — Deployment Commands"
@@ -23,6 +23,8 @@ help:
@echo " make check Check deployment status"
@echo " make ssl Setup SSL on VPS"
@echo " make service Install/restart hermes-gateway service"
@echo " make metrics View crisis metrics summary"
@echo " make metrics-json Export crisis metrics as JSON"
@echo ""
deploy:
@@ -46,3 +48,9 @@ ssl:
service:
ssh root@$(VPS) "cd /opt/the-door && bash deploy/deploy.sh --service"
metrics:
python3 -m crisis.metrics --summary
metrics-json:
python3 -m crisis.metrics --json

View File

@@ -8,6 +8,7 @@ from .detect import detect_crisis, CrisisDetectionResult, format_result, get_urg
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
from .metrics import CrisisMetrics, AggregateMetrics
__all__ = [
"detect_crisis",
@@ -23,4 +24,6 @@ __all__ = [
"CrisisSessionTracker",
"SessionState",
"check_crisis_with_session",
"CrisisMetrics",
"AggregateMetrics",
]

View File

@@ -104,9 +104,13 @@ MEDIUM_INDICATORS = [
r"\blost\s+all\s+hope\b",
r"\bno\s+tomorrow\b",
# Contextual versions (from crisis_detector.py legacy)
# Keep only medium-only patterns here; stronger overlaps live in HIGH_INDICATORS.
r"\bfeel(?:s|ing)?\s+(?:so\s+)?worthless\b",
r"\bfeel(?:s|ing)?\s+(?:so\s+)?hopeless\b",
r"\bfeel(?:s|ing)?\s+trapped\b",
r"\bfeel(?:s|ing)?\s+desperate\b",
r"\bno\s+future\s+(?:for\s+me|ahead|left)\b",
r"\bnothing\s+left\s+(?:to\s+(?:live|hope)\s+for|inside)\b",
r"\bgive(?:n)?\s*up\s+on\s+myself\b",
]
LOW_INDICATORS = [

244
crisis/metrics.py Normal file
View File

@@ -0,0 +1,244 @@
"""
crisis/metrics.py — Aggregate crisis detection metrics.
Tracks session-level crisis data for aggregate reporting.
Privacy-first: stores only aggregate counts, never user content.
Usage:
from crisis.metrics import CrisisMetrics
metrics = CrisisMetrics()
metrics.record_session(tracker.state)
summary = metrics.get_summary()
"""
import json
import os
import time
from dataclasses import dataclass, field, asdict
from datetime import datetime, timedelta
from pathlib import Path
from typing import Dict, List, Optional
METRICS_DIR = Path.home() / ".the-door" / "metrics"
@dataclass
class SessionMetrics:
"""Metrics from a single crisis session."""
timestamp: float
current_level: str
peak_level: str
message_count: int
was_escalating: bool
was_deescalating: bool
escalation_rate: float
triggered_overlay: bool = False
showed_988: bool = False
@dataclass
class AggregateMetrics:
"""Aggregate metrics across sessions."""
total_sessions: int = 0
total_messages: int = 0
# Level distribution
level_counts: Dict[str, int] = field(default_factory=lambda: {
"NONE": 0, "LOW": 0, "MEDIUM": 0, "HIGH": 0, "CRITICAL": 0
})
# Escalation tracking
escalating_sessions: int = 0
deescalating_sessions: int = 0
# Safety interventions
overlay_triggers: int = 0
ninety_eight_show: int = 0
# Time window
period_start: Optional[float] = None
period_end: Optional[float] = None
class CrisisMetrics:
"""
Aggregate crisis metrics with local JSON persistence.
Privacy-first: stores only aggregate counts per day.
Never stores user messages, content, or identifying info.
"""
def __init__(self, metrics_dir: Optional[Path] = None):
self.metrics_dir = metrics_dir or METRICS_DIR
self.metrics_dir.mkdir(parents=True, exist_ok=True)
self._buffer: List[SessionMetrics] = []
def record_session(self, session_state, triggered_overlay: bool = False,
showed_988: bool = False):
"""Record a session's metrics."""
from .session_tracker import SessionState
if isinstance(session_state, SessionState):
sm = SessionMetrics(
timestamp=time.time(),
current_level=session_state.current_level,
peak_level=session_state.peak_level,
message_count=session_state.message_count,
was_escalating=session_state.is_escalating,
was_deescalating=session_state.is_deescalating,
escalation_rate=session_state.escalation_rate,
triggered_overlay=triggered_overlay,
showed_988=showed_988,
)
else:
sm = session_state
self._buffer.append(sm)
self._flush()
def _flush(self):
"""Write buffered sessions to daily file."""
if not self._buffer:
return
today = datetime.utcnow().strftime("%Y-%m-%d")
filepath = self.metrics_dir / f"{today}.jsonl"
with open(filepath, 'a') as f:
for sm in self._buffer:
f.write(json.dumps(asdict(sm)) + '\n')
self._buffer.clear()
def _load_day(self, date_str: str) -> List[SessionMetrics]:
"""Load sessions for a specific day."""
filepath = self.metrics_dir / f"{date_str}.jsonl"
if not filepath.exists():
return []
sessions = []
with open(filepath) as f:
for line in f:
if line.strip():
data = json.loads(line)
sessions.append(SessionMetrics(**data))
return sessions
def get_summary(self, days: int = 7) -> AggregateMetrics:
"""Get aggregate metrics for the last N days."""
agg = AggregateMetrics()
now = datetime.utcnow()
for i in range(days):
date = (now - timedelta(days=i)).strftime("%Y-%m-%d")
sessions = self._load_day(date)
for sm in sessions:
agg.total_sessions += 1
agg.total_messages += sm.message_count
# Level counts (use peak level)
level = sm.peak_level
agg.level_counts[level] = agg.level_counts.get(level, 0) + 1
if sm.was_escalating:
agg.escalating_sessions += 1
if sm.was_deescalating:
agg.deescalating_sessions += 1
if sm.triggered_overlay:
agg.overlay_triggers += 1
if sm.showed_988:
agg.ninety_eight_show += 1
# Time window
if agg.period_start is None or sm.timestamp < agg.period_start:
agg.period_start = sm.timestamp
if agg.period_end is None or sm.timestamp > agg.period_end:
agg.period_end = sm.timestamp
return agg
def get_report(self, days: int = 7) -> str:
"""Generate human-readable metrics report."""
agg = self.get_summary(days)
lines = []
lines.append("=" * 50)
lines.append(" CRISIS METRICS REPORT")
lines.append(f" Last {days} days")
if agg.period_start:
start = datetime.fromtimestamp(agg.period_start).strftime("%Y-%m-%d %H:%M")
lines.append(f" Period: {start} → now")
lines.append("=" * 50)
lines.append(f"\n Sessions: {agg.total_sessions}")
lines.append(f" Messages tracked: {agg.total_messages}")
lines.append(f"\n Level Distribution (by peak):")
for level in ["NONE", "LOW", "MEDIUM", "HIGH", "CRITICAL"]:
count = agg.level_counts.get(level, 0)
pct = (count / agg.total_sessions * 100) if agg.total_sessions > 0 else 0
bar = "" * int(pct / 5)
lines.append(f" {level:<10} {count:>5} ({pct:>5.1f}%) {bar}")
lines.append(f"\n Escalations: {agg.escalating_sessions}")
lines.append(f" De-escalations: {agg.deescalating_sessions}")
lines.append(f" Overlay triggers: {agg.overlay_triggers}")
lines.append(f" 988 shown: {agg.ninety_eight_show}")
if agg.total_sessions > 0:
escalation_rate = agg.escalating_sessions / agg.total_sessions * 100
lines.append(f"\n Escalation rate: {escalation_rate:.1f}%")
lines.append("=" * 50)
return "\n".join(lines)
def get_json(self, days: int = 7) -> str:
"""Export metrics as JSON."""
agg = self.get_summary(days)
return json.dumps(asdict(agg), indent=2)
def main():
"""CLI entry point for crisis metrics."""
import argparse
parser = argparse.ArgumentParser(description="Crisis Detection Metrics")
parser.add_argument("--summary", action="store_true", help="Show summary report")
parser.add_argument("--json", action="store_true", help="JSON export")
parser.add_argument("--days", type=int, default=7, help="Days to include")
parser.add_argument("--demo", action="store_true", help="Generate demo data")
args = parser.parse_args()
metrics = CrisisMetrics()
if args.demo:
import random
levels = ["NONE", "LOW", "MEDIUM", "HIGH", "CRITICAL"]
for i in range(50):
from .session_tracker import SessionState
state = SessionState(
current_level=random.choice(levels),
peak_level=random.choice(levels),
message_count=random.randint(1, 20),
is_escalating=random.random() > 0.7,
is_deescalating=random.random() > 0.8,
escalation_rate=random.random(),
)
metrics.record_session(
state,
triggered_overlay=random.random() > 0.8,
showed_988=random.random() > 0.7,
)
print("Generated 50 demo sessions.")
if args.json:
print(metrics.get_json(args.days))
else:
print(metrics.get_report(args.days))
if __name__ == "__main__":
main()

View File

@@ -680,7 +680,7 @@ html, body {
<!-- Footer -->
<footer id="footer">
<a href="/about.html" aria-label="About The Door">about</a>
<a href="/about" aria-label="About The Door">about</a>
<button id="safety-plan-btn" aria-label="Open My Safety Plan">my safety plan</button>
<button id="clear-chat-btn" aria-label="Clear chat history">clear chat</button>
</footer>

View File

@@ -0,0 +1,118 @@
"""
Tests for crisis/metrics.py — Aggregate crisis metrics.
"""
import json
import os
import shutil
import tempfile
import unittest
from pathlib import Path
import sys
sys.path.insert(0, str(Path(__file__).parent.parent))
from crisis.metrics import CrisisMetrics, SessionMetrics, AggregateMetrics
class TestCrisisMetrics(unittest.TestCase):
def setUp(self):
self.tmpdir = tempfile.mkdtemp()
self.metrics = CrisisMetrics(Path(self.tmpdir))
def tearDown(self):
shutil.rmtree(self.tmpdir)
def test_record_session_creates_file(self):
sm = SessionMetrics(
timestamp=1700000000,
current_level="LOW",
peak_level="MEDIUM",
message_count=5,
was_escalating=True,
was_deescalating=False,
escalation_rate=0.5,
)
self.metrics.record_session(sm)
files = list(Path(self.tmpdir).glob("*.jsonl"))
self.assertEqual(len(files), 1)
def test_record_session_writes_jsonl(self):
sm = SessionMetrics(
timestamp=1700000000,
current_level="HIGH",
peak_level="CRITICAL",
message_count=10,
was_escalating=True,
was_deescalating=False,
escalation_rate=1.0,
triggered_overlay=True,
showed_988=True,
)
self.metrics.record_session(sm)
files = list(Path(self.tmpdir).glob("*.jsonl"))
with open(files[0]) as f:
data = json.loads(f.readline())
self.assertEqual(data['peak_level'], 'CRITICAL')
self.assertTrue(data['triggered_overlay'])
def test_get_summary_empty(self):
agg = self.metrics.get_summary(days=7)
self.assertEqual(agg.total_sessions, 0)
self.assertEqual(agg.total_messages, 0)
def test_get_summary_with_data(self):
for level in ["LOW", "MEDIUM", "HIGH"]:
sm = SessionMetrics(
timestamp=1700000000,
current_level=level,
peak_level=level,
message_count=3,
was_escalating=level != "LOW",
was_deescalating=False,
escalation_rate=0.5,
)
self.metrics.record_session(sm)
agg = self.metrics.get_summary(days=1)
self.assertEqual(agg.total_sessions, 3)
self.assertEqual(agg.total_messages, 9)
self.assertEqual(agg.escalating_sessions, 2)
def test_get_report_returns_string(self):
sm = SessionMetrics(
timestamp=1700000000,
current_level="LOW",
peak_level="LOW",
message_count=5,
was_escalating=False,
was_deescalating=False,
escalation_rate=0.0,
)
self.metrics.record_session(sm)
report = self.metrics.get_report(days=1)
self.assertIn("CRISIS METRICS REPORT", report)
self.assertIn("Sessions:", report)
def test_get_json_returns_valid(self):
sm = SessionMetrics(
timestamp=1700000000,
current_level="MEDIUM",
peak_level="MEDIUM",
message_count=3,
was_escalating=False,
was_deescalating=False,
escalation_rate=0.0,
)
self.metrics.record_session(sm)
json_str = self.metrics.get_json(days=1)
data = json.loads(json_str)
self.assertEqual(data['total_sessions'], 1)
if __name__ == "__main__":
unittest.main()

View File

@@ -1,50 +0,0 @@
"""Regression tests for duplicate crisis indicators cleanup (#123)."""
import os
import re
import sys
import unittest
from pathlib import Path
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from crisis.detect import detect_crisis
ROOT = Path(__file__).resolve().parents[1]
DETECT_FILE = ROOT / "crisis" / "detect.py"
class TestCrisisIndicatorDedup(unittest.TestCase):
def _extract_indicator_list(self, name: str) -> set[str]:
text = DETECT_FILE.read_text(encoding="utf-8")
match = re.search(rf"{name}\s*=\s*\[(.*?)\n\]", text, re.S)
self.assertIsNotNone(match, f"{name} list missing")
return {
line.strip().rstrip(",")
for line in match.group(1).splitlines()
if line.strip().startswith('r"')
}
def test_high_and_medium_indicator_lists_do_not_overlap(self):
high = self._extract_indicator_list("HIGH_INDICATORS")
medium = self._extract_indicator_list("MEDIUM_INDICATORS")
overlap = high & medium
self.assertEqual(set(), overlap, f"duplicate cross-tier patterns found: {sorted(overlap)}")
def test_removed_duplicates_still_classify_at_high_tier(self):
cases = {
"I feel hopeless": "HIGH",
"I feel trapped": "HIGH",
"I feel desperate": "HIGH",
"I have no future ahead": "HIGH",
"I have given up on myself": "HIGH",
}
for text, expected in cases.items():
with self.subTest(text=text):
result = detect_crisis(text)
self.assertEqual(expected, result.level)
if __name__ == "__main__":
unittest.main()