Skip to main content

claim_synthesis

Summary synthesis from claim hierarchies.

This module provides functions for generating coherent narrative summaries from structured claim hierarchies. Supports:

  • Single-source synthesis (one video's claims)
  • Multi-source synthesis (multiple videos' claims for collections)
  • Conflict detection and resolution
  • Hierarchical claim structure preservation
  • Claim relation integration (supports, conflicts, etc.)

logging​

Any​

GenerationConfig​

LLMLoader​

ClaimRelationship​

ClaimSource​

logger​

synthesize_summary_from_claims​

async def synthesize_summary_from_claims(
        claim_sources: list[ClaimSource],
        claim_relations: list[ClaimRelationship] | None,
        synthesis_strategy: str, ontology_context: dict[str, Any] | None,
        persona_context: dict[str, Any] | None, llm_loader: LLMLoader,
        max_length: int, include_conflicts: bool,
        include_citations: bool) -> list[dict[str, Any]]

Synthesize narrative summary from claim hierarchies.

Parameters​

claim_sources : list[ClaimSource] Claim hierarchies from one or more videos/collections. claim_relations : list[ClaimRelationship] | None Relationships between claims (conflicts, support, etc.). synthesis_strategy : str Strategy: "hierarchical", "chronological", "narrative", "analytical". ontology_context : dict[str, Any] | None Ontology types and glosses for context. persona_context : dict[str, Any] | None Persona information for perspective. llm_loader : LLMLoader Loaded LLM for generation. max_length : int Maximum summary length in words. include_conflicts : bool Whether to explicitly mention conflicts. include_citations : bool Whether to include claim citations.

Returns​

list[dict[str, Any]] Summary as GlossItem array with # and @ references.

build_synthesis_prompt​

def build_synthesis_prompt(claim_sources: list[ClaimSource],
                           claim_relations: list[ClaimRelationship] | None,
                           synthesis_strategy: str,
                           ontology_context: dict[str, Any] | None,
                           persona_context: dict[str, Any] | None,
                           max_length: int, include_conflicts: bool,
                           include_citations: bool) -> str

Build LLM prompt for summary synthesis.

Parameters​

claim_sources : list[ClaimSource] Claim hierarchies from sources. claim_relations : list[ClaimRelationship] | None Relationships between claims. synthesis_strategy : str Synthesis strategy. ontology_context : dict[str, Any] | None Ontology types and glosses. persona_context : dict[str, Any] | None Persona information. max_length : int Maximum summary length. include_conflicts : bool Explicitly mention conflicts. include_citations : bool Include citations.

Returns​

str Formatted prompt for LLM.