Synthesis Rules Schema

[mathaix]

User Story

As a developer
I want the SynthesisBlueprint schema defined
So that the synthesis pipeline knows how to process extracted entities

Acceptance Criteria

SynthesisBlueprint Structure

class SynthesisBlueprint(BaseModel):
    entity_resolution_rules: list[ResolutionRule]
    correlation_rules: list[CorrelationRule]
    analysis_frameworks: list[AnalysisFramework]
    conflict_resolution: ConflictResolutionSpec

ResolutionRule

Entity resolution determines when two extracted entities are the same:

• entity_type: Which entity type this rule applies to
• matching_fields: Fields to compare (e.g., ["name", "vendor"])
• similarity_threshold: 0-1 threshold for match (default 0.85)
• merge_strategy: first | most_recent | most_confident | manual
• blocking_keys: Fields used for blocking to reduce comparisons

CorrelationRule

Cross-interview pattern detection:

• name: Rule identifier
• description: What pattern this detects
• entities_involved: Entity types involved
• correlation_logic: How to detect (cypher or description)
• output_type: What this produces (insight, pattern, conflict)
• min_sources: Minimum interviews mentioning to trigger

AnalysisFramework

Domain-specific analysis:

• name: Framework name (e.g., "overlap_analysis", "risk_matrix")
• description: What this framework does
• inputs: Entity types needed
• algorithm: How to compute (cypher, python, or description)
• output_schema: Structure of output
• visualization: How to visualize (matrix, chart, diagram)

ConflictResolutionSpec

When interviewees disagree:

• detection_threshold: Similarity to flag as conflict
• resolution_strategy: majority | expertise_weighted | flag_for_review
• expert_roles: Roles whose input has higher weight
• escalation_threshold: Conflicts above this escalate to manager

Standard Analysis Frameworks

Pre-built frameworks by project type:

• M&A: System overlap matrix, integration complexity, org alignment
• ERP: Capability gap analysis, integration inventory, data quality
• Process: Process variance analysis, bottleneck detection, automation potential
• Compliance: Control coverage, gap severity, remediation priority

Validation

• Entity types referenced exist in extraction schema
• Merge strategies are valid enum values
• Algorithm references are parseable
• Output schemas are valid JSON schemas

Technical Notes

• Resolution rules drive entity merge in Neo4j
• Correlation rules become Cypher queries
• Frameworks reference ANALYSIS-TEMPLATES.md

Definition of Done

• All synthesis models defined
• Standard frameworks implemented
• Validation complete
• Integration with synthesis pipeline documented
• Code reviewed and merged

---

🤖 Generated with Claude Code