Upload schemas.py with huggingface_hub

schemas.py

@@ -0,0 +1,126 @@
+"""State schema and Pydantic output models for the ITIL Reflexion Agent."""
+
+from typing import TypedDict, List, Dict, Annotated, Optional
+from pydantic import BaseModel, Field
+import operator
+
+
+# =============================================================================
+# LangGraph State
+# =============================================================================
+
+class RFCState(TypedDict):
+    """State object passed between LangGraph nodes."""
+    # Input data
+    scenario_id: str
+    incidents: List[Dict]
+    cmdb_info: Dict
+    scenario_meta: Dict
+    custom_data: Optional[Dict]  # User-uploaded incidents/CMDB data
+
+    # Iteration control
+    iteration: int
+    max_iterations: int
+    score_threshold: int
+
+    # Agent outputs
+    rfc: str
+    critique: Dict
+    feedback: str
+    prompt_strategy: str
+    improvement_pattern: str
+
+    # Accumulated history (append-only via operator.add)
+    history: Annotated[List[Dict], operator.add]
+
+    # Control flow
+    should_continue: bool
+    final_result: Optional[Dict]
+    cab_summary: str
+
+    # Streaming
+    stream_queue: Optional[object]
+
+
+# =============================================================================
+# Pydantic Output Models (for structured LLM output)
+# =============================================================================
+
+class RFCScores(BaseModel):
+    """Six-dimension scoring for an RFC iteration."""
+    overall_quality: float = Field(ge=0, le=10, description="Overall RFC quality 0-10")
+    itil_compliance: float = Field(ge=0, le=10, description="ITIL framework compliance 0-10")
+    risk_level: float = Field(ge=0, le=10, description="Risk level 0-10 (lower is better)")
+    business_value: float = Field(ge=0, le=10, description="Business value articulation 0-10")
+    technical_readiness: float = Field(ge=0, le=10, description="Technical readiness 0-10")
+    stakeholder_confidence: float = Field(ge=0, le=10, description="Stakeholder confidence 0-10")
+
+
+class ExecutiveSummary(BaseModel):
+    """Executive summary for CAB review."""
+    recommendation: str = Field(description="e.g. CONDITIONAL APPROVAL, APPROVED FOR PRODUCTION")
+    deployment_risk: str = Field(description="e.g. HIGH, MEDIUM, LOW")
+    business_impact: str = Field(description="Business impact level")
+    cab_approval_probability: float = Field(ge=0, le=1, description="Probability 0-1")
+    estimated_roi: str = Field(description="ROI estimate")
+    key_concerns: List[str] = Field(default_factory=list, description="Remaining concerns")
+
+
+class RFCSummary(BaseModel):
+    """Structured RFC summary."""
+    title: str
+    objective: str
+    business_justification: str
+    technical_approach: str
+    rollback_plan_status: str
+    testing_status: str
+    timeline: str
+    impact: str
+
+
+class CriticalIssue(BaseModel):
+    """A critical issue identified in the RFC."""
+    issue: str
+    category: str
+    severity: str = Field(description="HIGH, MEDIUM, or LOW")
+    priority: str
+    impact: str
+
+
+class Improvement(BaseModel):
+    """A recommended improvement."""
+    action: str
+    priority: str = Field(description="CRITICAL, HIGH, MEDIUM, or LOW")
+    estimated_impact: str
+    effort_hours: float
+
+
+class ChangeCategoryScore(BaseModel):
+    """Score for a change category."""
+    score: float = Field(ge=0, le=10)
+    status: str = Field(description="EXCELLENT, GOOD, ADEQUATE, or NEEDS IMPROVEMENT")
+
+
+class ChangeCategories(BaseModel):
+    """Four change management categories."""
+    technical: ChangeCategoryScore
+    procedural: ChangeCategoryScore
+    compliance: ChangeCategoryScore
+    communication: ChangeCategoryScore
+
+
+class EvaluationOutput(BaseModel):
+    """Structured output from the evaluator agent."""
+    scores: RFCScores
+    executive_summary: ExecutiveSummary
+    rfc_summary: RFCSummary
+    critical_issues: List[CriticalIssue] = Field(default_factory=list)
+    improvements: List[Improvement] = Field(default_factory=list)
+    change_categories: ChangeCategories
+
+
+class ReflectionOutput(BaseModel):
+    """Structured output from the reflector agent."""
+    feedback: str = Field(description="Actionable feedback for the actor")
+    focus_areas: List[str] = Field(description="Key areas to improve")
+    strategy_recommendation: str = Field(description="Suggested prompt strategy for next iteration")