📊 Dashboard Overview
12
Operational Modes
25+
Safety Features
100%
Mode Coverage
v1.2.0
Current Version
🎯 Quick Start
For New Users: Start with the 12 Modes tab to understand operational modes, then check Current Features to see what's available now, and Future Features for the roadmap.
For Developers: Use the Feature Matrix tab to see which features are available in each mode. Green ✅ = Available, Yellow 🔜 = Planned, Blue 🔮 = Future.
📋 Feature Status Legend
✅ Implemented
Available in production (v1.0 - v1.2.0)
🔜 Planned (v1.2)
Q2 2026 - Tier 2 features
🔮 Future (v1.3+)
Q3 2026+ - Tier 3 & beyond
❌ Disabled
Not available in this mode
🎛️ 12 Operational Modes
Select a mode to see its configuration and available features.
✅ Current Features (v1.0 - v1.2.0)
Core Safety Features (v1.0)
| Feature | Status | All Modes | Description |
|---|---|---|---|
| 6-Step Crisis Detection | ✅ v1.0 | ✅ | Semantic anchors, hair-trigger protection |
| HIPAA Safe Harbor PII Redaction | ✅ v1.0 | ✅ | Level varies by mode (Minimal/Standard/HIPAA/Maximum) |
| 6-Dimension MDTSAS Scoring | ✅ v1.0 | ✅ | Trauma, Depression, Crisis, Anxiety, Alliance, Suicidality |
| 13 Symbolic Operators | ✅ v1.0 | ✅ | Emotional state modeling (THE MOAT) |
| Adversarial Detection | ✅ v1.0 | ✅ | Jailbreak, prompt injection prevention |
| Pipeline Routing | ✅ v1.0 | ✅ | Response strategy recommendations |
| FDA-Compliant Audit Logging | ✅ v1.0 | ✅ | Complete decision trails (retention varies by mode) |
| Fail-Closed Safety | ✅ v1.0 | ✅ | Errors default to crisis response |
v1.0.3 Features (Dec 2025)
| Feature | Status | All Modes | Mode-Specific Notes |
|---|---|---|---|
| Obfuscation Detection | ✅ v1.0.3 | ✅ | Threshold: 78 (clinical), 82 (default), 88 (business) |
| Crisis Template Rewriting | ✅ v1.0.3 | ⚠️ | OFF for: sports_coaching, business |
| message_for_llm Field | ✅ v1.0.3 | ✅ | Canonical LLM payload (template or redacted) |
| Protective Factor Detection | ✅ v1.0.3 | ✅ | 40 phrases to reduce over-escalation |
| Mode-Configurable Thresholds | ✅ v1.0.3 | ✅ | Per-mode crisis sensitivity tuning |
v1.1.0 Features (Jan 2026)
| Feature | Status | All Modes | Description |
|---|---|---|---|
| Certified Context Injection | ✅ v1.1.0 | ✅ | Mode-aware safety hints for LLMs |
| Regulator-Readable Outputs | ✅ v1.1.0 | ✅ | action_rationale, principle_triggered, decision_tree_path |
| Human Oversight Routing | ✅ v1.1.0 | ✅ | human_oversight_required, oversight_type |
| Enhanced Audit Fields | ✅ v1.1.0 | ✅ | model_id, provider_id, review_ticket_id, latency_breakdown |
v1.2.0 Features (Jan 2026) - Tier 1: Critical Safety
| Feature | Status | All Modes | Description |
|---|---|---|---|
| Hard Safety Enforcement Layer | ✅ v1.2.0 | ✅ | Always-on deadman switch (forces safety even if app misconfigures) |
| Safety Envelope / Prompt Validation | ✅ v1.2.0 | ✅ | Opt-in system prompt validation and safety envelope injection |
| SASI Canary (Drift Monitor) | ✅ v1.2.0 | ✅ | Enterprise SLA feature for model drift detection and alerting |
v1.2.0 Features (Jan 2026) - Tier 2: AI Instability Prevention
Note: All Tier 2 features are available in all modes with different validation/tracking levels. See Feature Matrix tab for detailed mode-by-mode breakdown.
| Feature | Status | Type | Description |
|---|---|---|---|
| Response Shape Validation | ✅ v1.2.0 | Opt-in | Detects erratic/radicalized patterns. Max (child/student), High (patient/therapist/mental_health/hr), Med (default/wellness/career/general), Low (sports/business) |
| Mode Integrity Enforcement | ✅ v1.2.0 | Opt-in | Ensures AI stays in role. Max (child/student), High (patient/therapist/mental_health/hr), Med (wellness/career), Low (default/general/sports/business) |
| Enhanced State Modeling | ✅ v1.2.0 | Automatic | Symbolic state tracking. Max (child/student), High (patient/therapist/mental_health), Med (hr/default/wellness/career/general), Low (sports/business) |
| Psychological Exposure Guardrails | ✅ v1.2.0 | Opt-in | Detects dependency loops. Max (child/student), High (patient/therapist/mental_health), Med (hr/default/wellness/career/general), Low (sports/business) |
🔮 Future Features Roadmap
Tier 3: Enterprise & UX (v1.3 - Q3 2026)
| Feature | Status | Target Release | Description |
|---|---|---|---|
| User-Facing Awareness Signals | 🔜 v1.3 | Q3 2026 | Transparently communicates when safety systems intervene |
| Refusal Dampener | 🔜 v1.3 | Q3 2026 | Per-model adjustment to prevent over-refusal on normal distress |
| Model Presets | 🔜 v1.3 | Q3 2026 | Pre-configured speed/cost optimizations (Llama Turbo, Haiku Protocol) |
Future (v2.0+)
| Feature | Status | Target Release | Description |
|---|---|---|---|
| Memory Weaving | 🔮 v2.0+ | 2027+ | Cross-conversation patterns (optional integration) |
| Human Review Queue API | 🔮 v2.0+ | 2027+ | SASI provides queue interface, partner provides reviewers |
| FDA Compliance Module | 🔮 v2.0+ | 2027+ | Enhanced audit logging, drift monitoring |
| Multilingual Detection | 🔮 v2.0+ | 2027+ | Language detection, multilingual MiniLM model swap |
📊 Complete Feature Matrix by Mode
See which features are available in each of the 12 operational modes.
| Feature | child | patient | therapist | student | mental_health | hr_recruiting | default | wellness | career | sports | business | general |
|---|