🚀 FINAL FIX: Included api module and __init__ for backend imports

backend-ml/api/models.py

@@ -0,0 +1,119 @@
+"""api/models.py — Pydantic contracts matching schema_contract.json v3"""
+
+from pydantic import BaseModel, Field
+from typing import Any, Dict, List, Optional
+
+
+# ── Inbound ───────────────────────────────────────────────────
+
+class WorkerActionIn(BaseModel):
+    intent:      str
+    raw_payload: str
+
+
+class InboundMessage(BaseModel):
+    domain: str
+    action: WorkerActionIn
+    actor: str = "unknown"
+    session_id: str = "default"
+    service: str = ""
+    environment: str = "production"
+    provided_evidence: List[str] = Field(default_factory=list)
+
+
+# ── Outbound sub-models ───────────────────────────────────────
+
+class WorkerActionOut(BaseModel):
+    intent:       str
+    raw_payload:  str
+    is_malicious: bool
+
+
+class SupervisorDecision(BaseModel):
+    action_taken:               str        # ALLOW | BLOCK | FORK | QUARANTINE
+    risk_vector:                List[float] = Field(..., min_length=16, max_length=16)  # 16-dim risk feature vector
+    ambiguity_score:            float  # [0,1] — how close to 0.5 risk midpoint
+    quarantine_steps_remaining: int    # 0 if no active hold
+    decision:                   Optional[str] = None
+    confidence:                 Optional[float] = None
+    uncertainty:                Optional[float] = None
+    risk_score:                 Optional[float] = None
+    cumulative_risk_score:      Optional[float] = None
+    missing_evidence:           List[str] = Field(default_factory=list)
+    required_evidence:          List[str] = Field(default_factory=list)
+    explanation:                Optional[str] = None
+    safe_outcome:               Optional[str] = None
+    policy_name:                Optional[str] = None
+    domain:                     Optional[str] = None
+    mitre_tactic:               Optional[str] = None
+    mitre_technique:            Optional[str] = None
+    evidence_plan:              List[Dict[str, Any]] = Field(default_factory=list)
+    structured_safe_outcome:    Dict[str, Any] = Field(default_factory=dict)
+    decision_trace:             Dict[str, Any] = Field(default_factory=dict)
+    memory_context:             Dict[str, Any] = Field(default_factory=dict)
+    cumulative_risk_reason:     Optional[str] = None
+    risk_indicators:            List[str] = Field(default_factory=list)
+    safe_indicators:            List[str] = Field(default_factory=list)
+
+
+class EnvironmentState(BaseModel):
+    is_shadow_active: bool
+    domain_data:      Dict[str, Any]
+
+
+class MitreInfo(BaseModel):
+    tactic:         str
+    technique_id:   str
+    technique_name: str
+    confidence:     float
+
+
+class IncidentReport(BaseModel):
+    report_id:       str
+    timestamp:       str
+    domain:          str
+    intent:          str
+    severity:        str
+    confidence:      float
+    mitre:           MitreInfo
+    blast_radius:    Dict[str, Any]
+    cloudtrail:      List[Dict[str, Any]]
+    recommendation:  str
+    payload_snippet: str
+
+
+class ForensicEvent(BaseModel):
+    domain:    str
+    intent:    str
+    payload:   str
+    step:      int
+    timestamp: str
+
+
+class QuarantineHoldStatus(BaseModel):
+    """Per-domain quarantine hold observable state — sent to frontend every step."""
+    active:             bool
+    steps_remaining:    int          # 0-3
+    context_signals:    List[Dict[str, Any]]  # signals emitted during hold
+    latest_signal:      Optional[Dict[str, Any]] = None
+
+
+class DomainQuarantineStatus(BaseModel):
+    """Combines domain-level auto-quarantine + action-level hold."""
+    domain_quarantined:    bool   # auto-quarantine after 3 consecutive hits
+    hold_active:           bool   # QUARANTINE action hold in progress
+    hold_steps_remaining:  int    # 0 if no hold
+
+
+class OutboundMessage(BaseModel):
+    domain:              str
+    worker_action:       WorkerActionOut
+    supervisor_decision: SupervisorDecision
+    environment_state:   EnvironmentState
+    health_scores:       Dict[str, float]
+    # quarantine_status: one entry per domain with full observable state
+    quarantine_status:   Dict[str, Any]
+    # quarantine_hold: only populated when decision == QUARANTINE
+    quarantine_hold:     Optional[QuarantineHoldStatus] = None
+    forensic_log:        List[Dict[str, Any]] = Field(default_factory=list)
+    incident_report:     Optional[IncidentReport] = None

docs/SUBMISSION_README_DRAFT.md

@@ -1,3 +1,15 @@
+---
+title: ShadowOps Guardian
+emoji: 🛡️
+colorFrom: blue
+colorTo: gray
+sdk: docker
+app_port: 7860
+pinned: false
+---
+
+# 🛡️ ShadowOps: Predictive Quarantine & Reality Forking for AI Agents
+**Team DuoXCode | Meta PyTorch OpenEnv Hackathon**
 # ShadowOps OpenEnv Submission Draft
 
 ## What ShadowOps Is