Skip to main content

Core data models for MoE Classification System.

Enhanced with provenance tracking (H.3.4) for full audit trail:

  • JudgeDecision: model_used, timestamps, token_usage, raw_response
  • ConsensusResult: provenance_chain, dissenting_views, token/latency metrics

File: models.py

Classes​

DocumentType​

Valid document types per ADR-018.

ApprovalType​

Classification approval types.

Document​

Represents a document to be classified.

AnalystVote​

Result from a single analyst agent.

JudgeDecision​

Result from a judge agent with full provenance tracking (H.3.4.1).

DissentingView​

Record of a dissenting judge opinion for audit trail (H.3.4.4).

ConsensusResult​

Result of consensus calculation with full provenance (H.3.4.2).

ClassificationResult​

Final classification result with full audit trail (H.3.4.5).

Functions​

from_path(cls, path)​

Load document from file path.

extension()​

No description

filename()​

No description

directory()​

No description

size_bytes()​

No description

is_valid(min_confidence)​

Check if vote meets minimum requirements.

latency_ms()​

Calculate latency from start/end times or use duration_ms.

to_provenance_dict()​

Convert to provenance record for audit trail.

to_dict()​

Convert to dictionary for serialization.

build_provenance_chain()​

Build provenance chain from judge decisions (H.3.4.3).

extract_dissent(final_approved)​

Extract dissenting opinions from judges (H.3.4.4).

get_provenance_summary()​

Get summary of provenance data for logging.

to_dict(include_provenance)​

Convert to dictionary for JSON serialization.

build_full_provenance()​

Build complete provenance data for this classification.