Events
Real-time audit log stream. Newest first. Trace IDs are clickable.
Timestamp
Trace ID
Feature
Actor
Action
Status
Algorithms
12 algorithms deployed. Click any card for drill-down + version history.
Errors
Last 24 hours · 3 active error types · circuit breakers nominal
Error rate (last 24h)
▼ 1.2% (down from 1.8% yesterday)
1.2%
Avg rate (24h)
52
Total errors (24h)
0
Circuit breakers tripped
100%
Audit log integrity
TOP ERROR TYPES (LAST 24H)
ClaudeApiTimeoutError21 eventslast: 23m ago
FirestorePermissionDenied18 eventslast: 1h ago
TwilioRateLimitError9 eventslast: 4h ago
PppClassifierLowConfidence4 eventslast: 6h ago
Participants
Coded IDs only. No PHI. Pre-BAA test mode. 12 enrolled, 3 active right now.
Feature Flags
Toggle features without redeploying. State changes are audit-logged.
Flag
Description
State
Last changed
Toggle
ppp_classifier_v1.2.3
PPP classifier — production version with phase confidence
ON · 100%
Bob · 2d ago
cascade_controller_v2
Cascade Controller with adaptive prioritization
ON · 100%
Bob · 3d ago
emotion_learning_loop
Cross-device emotion vocabulary learning (Candidate O)
COHORT · 50%
Bob · 4h ago
risk_anchor_awareness
Semantic risk-anchor detection (Candidate I)
ON · 100%
Bob · 6d ago
stage_conditioned_signaling
Phase-conditioned cascade tone (Candidate I.2)
ON · 100%
Bob · 6d ago
comparative_effectiveness
Outcomes comparison engine (Candidate Q)
OFF · 0%
Bob · 1d ago
haptic_first_wearable
Watch-first cascade delivery (Candidate J)
OFF · 0%
Bob · 12d ago
maintenance_mode
Emergency site-wide offline (use with care)
OFF · 0%
—
AI Assistant
Ask about anything in the audit log. Powered by Claude. Read-only — no autonomous actions.
BS
Walk me through what happened with
GBETA-D9E1-F7 this morning — I saw a cascade fired around 9 AM.✦
Yes — a clinically meaningful event. Let me walk through it. Full trace
09:14:02 UTC — Participant submitted morning check-in:
09:14:08 UTC —
09:14:11 UTC —
09:14:14 UTC — Guardian Ring received the cascade. Slot 1 (sponsor, coded "M.S.") contacted via SMS per the participant's cascade configuration.
09:21:43 UTC — 7 min 32 sec later — Slot 1 responded: "I'm here, calling you now." Cascade resolved. Chain of Custody anchored.
09:43:11 UTC — Participant submitted follow-up: "Talking with M.S. now. Better."
This is a clinical success: first-call resolution, sub-8-minute time to human contact, participant self-reported improvement within 30 minutes. Want me to draft a case-file note?
tr-d9e1f7-casc.
09:14:02 UTC — Participant submitted morning check-in:
"I keep thinking about calling M. today. Things have been tense at work and I don't think I can handle one more bad day without help."
09:14:05 UTC — risk-anchor-awareness v1.1.0 detected a high-salience pattern: name on the participant's own risk-anchor list co-occurring with Build-Up phase language markers ("don't think I can handle"). Risk score: 0.81.
09:14:08 UTC —
pppClassifyV1_2_3 classified the same text as Build-Up phase, confidence 0.79.
09:14:11 UTC —
cascadeV2 evaluated both signals together. Build-Up phase + active risk-anchor → triggered Considerate cascade mode per §4.2.
09:14:14 UTC — Guardian Ring received the cascade. Slot 1 (sponsor, coded "M.S.") contacted via SMS per the participant's cascade configuration.
09:21:43 UTC — 7 min 32 sec later — Slot 1 responded: "I'm here, calling you now." Cascade resolved. Chain of Custody anchored.
09:43:11 UTC — Participant submitted follow-up: "Talking with M.S. now. Better."
This is a clinical success: first-call resolution, sub-8-minute time to human contact, participant self-reported improvement within 30 minutes. Want me to draft a case-file note?
BS
Yes, please. Also — does this fit the pattern we've seen for this participant before?
✦
Good question. Let me check their history.
• Event 1 (day 8) — Build-Up + work-related risk anchor → cascade to sponsor → resolved in 12m 04s
• Event 2 (day 19) — Justification phase signals only → no cascade (correct per design)
• Event 3 (day 31) — Build-Up + relationship-related risk anchor → cascade to sponsor → resolved in 4m 18s
• Event 4 (today, day 41) — Build-Up + work-related risk anchor → cascade to sponsor → resolved in 7m 32s
Patterns I see: Cascade-resolved time has stayed sub-10-minute. Work-related risk anchors are the most frequent trigger. Sponsor responds reliably within 10 minutes. This is healthy engagement.
Suggestion for clinical review: if work stress is a recurring trigger for this participant, the Stage-Conditioned Signaling module (§4.7.9) could be tuned to their specific anchor pattern. That would be a per-participant tuning, not a global algorithm change. Want me to flag this for the next algorithm tuning cycle?
GBETA-D9E1-F7 has been on the platform 41 days. This is their 4th cascade event:
• Event 1 (day 8) — Build-Up + work-related risk anchor → cascade to sponsor → resolved in 12m 04s
• Event 2 (day 19) — Justification phase signals only → no cascade (correct per design)
• Event 3 (day 31) — Build-Up + relationship-related risk anchor → cascade to sponsor → resolved in 4m 18s
• Event 4 (today, day 41) — Build-Up + work-related risk anchor → cascade to sponsor → resolved in 7m 32s
Patterns I see: Cascade-resolved time has stayed sub-10-minute. Work-related risk anchors are the most frequent trigger. Sponsor responds reliably within 10 minutes. This is healthy engagement.
Suggestion for clinical review: if work stress is a recurring trigger for this participant, the Stage-Conditioned Signaling module (§4.7.9) could be tuned to their specific anchor pattern. That would be a per-participant tuning, not a global algorithm change. Want me to flag this for the next algorithm tuning cycle?
BS
Yes. And share this conversation with Dr. Dick at Rutgers — generate a snapshot so she can review it on her own time.
✦
Done. Flagged
Snapshot saved as
Share URL:
Audit log entry created:
GBETA-D9E1-F7 for the next algorithm tuning cycle with a note about work-related anchor specialization.
Snapshot saved as
snap_d9e1f7c1. Expires in 30 days. Contents: this full conversation, trace tr-d9e1f7-casc, the participant's last 41 days of cascade events, and the audit-anchored chain for today's cascade. Coded ID only — no PHI.
Share URL:
https://the-fires-within.com/guido-console/snapshot/snap_d9e1f7c1
Audit log entry created:
Bob Seibel created snap_d9e1f7c1 (recipient pending). When Dr. Dick views the snapshot, her identity will be appended to the audit log.
Control Arm
Matched controls, per-instrument survey cadences, missed-cadence alerts, and side-by-side outcome trajectories. Comparative Effectiveness Plan v1.1 §4 and §6.
Control-arm enrolled
34
target 30-60 · 1:1 to 1:2 matching
GUIDO arm enrolled
30
all matched · 4 awaiting controls
Survey completion (control)
87%
+4 pp vs. last week
Active missed-cadence alerts
2
1 high · 1 medium · fallback pending
PCL-5 Δ (study-level L4)
−6.2
d = 0.43 · 95% CI [0.12, 0.74]
Token-survey delivery rate
96%
SMS + email · single-use signed JWT
Missed-cadence alerts
· surveys overdue beyond ±2-day soft window · escalates to RA phone-call fallback
Participant
Issue
Fallback
Severity
CTRL-P-0042
Missed 2 consecutive biweekly AUDIT-C tokens (last submission day 12; 28 days elapsed).
RA phone-call scheduled
2026-06-19 · K. Nguyen High
2026-06-19 · K. Nguyen High
CTRL-P-0018
Missed weekly WHO-5 token (5 days overdue). Other instruments on cadence.
Auto-resend token
delivery confirmed Medium
delivery confirmed Medium
Side-by-side outcome trajectories
· L1 Instrument · L2 Domain · L3 Arm · L4 Study (Plan §7.1)
Instrument · Domain
GUIDO arm (n=30)
Control arm (n=34)
Δ effect
PCL-5 · PTSD domain
d=0.43
AUDIT-C · Substance domain
d=0.38
IDS-SR · Mood domain
d=0.21
WHO-5 · Well-being domain
d=0.47
L3 Arm · all domains combined
Z-composite −0.61
Z-composite −0.23
d=0.40
Control-arm participants — survey cadences
· per Plan §4.1 instrument-specific schedules · mode tags show how each survey landed
Completed on cadence
Pending (within window)
Fallback fired (in-clinic / RA call)
Missed (overdue)
CTRL-P-0042
matched to GBETA-D9E1-F7
PCL-5
3/3
AUDIT-C
3/5
DAST-10
2/3
WHO-5
8/8
at_home_token
in_clinic_tablet
phone_call
CTRL-P-0018
matched to GBETA-A4C8-B2
PCL-5
2/2
AUDIT-C
3/3
DAST-10
2/2
WHO-5
4/5
at_home_token
in_clinic_tablet
CTRL-P-0007
matched to GBETA-7F2A-C5
PCL-5
3/3
AUDIT-C
6/6
DAST-10
3/3
WHO-5
11/11
at_home_token
CTRL-P-0023
matched to GBETA-9B14-EE
PCL-5
1/2
AUDIT-C
2/2
DAST-10
1/2
WHO-5
3/3
at_home_token
Methodology · v1.1 enrichment.
Cadence per instrument follows Comparative Effectiveness Plan v1.1 §4.1. Collection cell —
(completed_by_role, completion_method) — is captured per record and modeled as a covariate per §4.2. Tokens are signed single-use JWTs; chain-of-custody is preserved across mode boundaries per §4.3. Composite effect sizes roll up Instrument → Domain → Arm → Study (L1 → L4) per §7.1. Pre-registered weights live in the Algorithms tab under effectiveness-rollup v1.1.0.