Skip to content

ontocast.agent.criticise_facts

Fact criticism agent for OntoCast.

This module provides functionality for analyzing and validating facts extracted from text chunks, ensuring their consistency and correctness.

criticise_facts(state, tools)

Analyze and validate facts in the current chunk.

This function performs a critical analysis of the facts in the current chunk, checking for consistency, correctness, and potential issues.

Parameters:

Name Type Description Default
state AgentState

The current agent state containing the chunk to analyze.

 required
tools ToolBox

The toolbox instance providing utility functions.

 required

Returns:

Name Type Description
AgentState AgentState

Updated state with analysis results.
Source code in ontocast/agent/criticise_facts.py 
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
def criticise_facts(state: AgentState, tools: ToolBox) -> AgentState:
    """Analyze and validate facts in the current chunk.

    This function performs a critical analysis of the facts in the current chunk,
    checking for consistency, correctness, and potential issues.

    Args:
        state: The current agent state containing the chunk to analyze.
        tools: The toolbox instance providing utility functions.

    Returns:
        AgentState: Updated state with analysis results.
    """
    if not state.current_chunk:
        logger.warning("No current chunk to analyze")
        return state

    logger.info("Criticize facts")

    llm_tool = tools.llm
    parser = PydanticOutputParser(pydantic_object=KGCritiqueReport)

    prompt = PromptTemplate(
        template=criticise_facts_prompt,
        input_variables=[
            "ontology",
            "document",
            "knowledge_graph",
            "format_instructions",
        ],
    )

    response = llm_tool(
        prompt.format_prompt(
            ontology=state.current_ontology.graph.serialize(format="turtle"),
            document=state.current_chunk.text,
            knowledge_graph=state.current_chunk.graph.serialize(format="turtle"),
            format_instructions=parser.get_format_instructions(),
        )
    )
    critique: KGCritiqueReport = parser.parse(response.content)
    logger.debug(
        f"Parsed critique report - success: {critique.facts_graph_derivation_success}, "
        f"score: {critique.facts_graph_derivation_score}"
    )

    if critique.facts_graph_derivation_success:
        logger.debug("Facts critique successful, clearing failure state")
        state.clear_failure()
    else:
        logger.debug("Facts critique failed, setting failure state")
        state.set_failure(
            stage=FailureStages.FACTS_CRITIQUE,
            reason=critique.facts_graph_derivation_critique_comment,
            success_score=critique.facts_graph_derivation_score,
        )
    return state