Skip to content

ontocast.stategraph.unit_context

Slim per-unit view of document state for the map-stage loops.

The unit render/critic loops only read a handful of document-level fields. Passing the full :class:~ontocast.onto.state.AgentState required a deep copy per unit (raw input bytes, docling document, every content unit, every graph), which dominated fan-out cost. UnitLoopContext carries exactly what the loops consume; large artifacts are shared by reference (read-only in the loops).

Classes

UnitLoopContext

Bases: BasePydanticModel

Document-level inputs for one unit's ontology/facts loop.

reduced_ontology_artifacts is shared by reference with the document state — the loops only read it. retrieval_metrics is per-unit and is folded back into the document state after the fan-out gathers.

Source code in ontocast/stategraph/unit_context.py 
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
class UnitLoopContext(BasePydanticModel):
    """Document-level inputs for one unit's ontology/facts loop.

    ``reduced_ontology_artifacts`` is shared by reference with the document
    state — the loops only read it. ``retrieval_metrics`` is per-unit and is
    folded back into the document state after the fan-out gathers.
    """

    model_config = ConfigDict(arbitrary_types_allowed=True)

    ontology_context_mode: OntologyContextMode
    ontology_selection_user_instruction: str = ""
    ontology_context_fixed_ontology_id: str = ""
    reduced_ontology_artifacts: list[Ontology] = Field(default_factory=list)
    budget_tracker: BudgetTracker = Field(default_factory=BudgetTracker)
    retrieval_metrics: dict[str, int | float | str | dict[str, Any]] = Field(
        default_factory=dict
    )

    @classmethod
    def from_agent_state(
        cls,
        state: AgentState,
        budget_tracker: BudgetTracker | None = None,
    ) -> UnitLoopContext:
        """Project the loop-relevant fields off a document state.

        Args:
            state: Document-level agent state.
            budget_tracker: Per-unit tracker LLM calls should be charged to;
                defaults to the document tracker (sequential pipelines).
        """
        return cls(
            ontology_context_mode=state.ontology_context_mode,
            ontology_selection_user_instruction=(
                state.ontology_selection_user_instruction
            ),
            ontology_context_fixed_ontology_id=(
                state.ontology_context_fixed_ontology_id
            ),
            reduced_ontology_artifacts=document_ontology_access(
                state
            ).reduced_artifacts(),
            budget_tracker=budget_tracker
            if budget_tracker is not None
            else state.budget_tracker,
        )

    def reduced_artifacts(self) -> list[Ontology]:
        """Reduced document ontology artifacts (shared, read-only)."""
        return list(self.reduced_ontology_artifacts)

Attributes

budget_tracker = Field(default_factory=BudgetTracker) class-attribute instance-attribute
model_config = ConfigDict(arbitrary_types_allowed=True) class-attribute instance-attribute
ontology_context_fixed_ontology_id = '' class-attribute instance-attribute
ontology_context_mode instance-attribute
ontology_selection_user_instruction = '' class-attribute instance-attribute
reduced_ontology_artifacts = Field(default_factory=list) class-attribute instance-attribute
retrieval_metrics = Field(default_factory=dict) class-attribute instance-attribute

Methods:

from_agent_state(state, budget_tracker=None) classmethod

Project the loop-relevant fields off a document state.

Parameters:

Name Type Description Default
state AgentState

Document-level agent state.

 required
budget_tracker BudgetTracker | None

Per-unit tracker LLM calls should be charged to; defaults to the document tracker (sequential pipelines).

 None
Source code in ontocast/stategraph/unit_context.py 
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
@classmethod
def from_agent_state(
    cls,
    state: AgentState,
    budget_tracker: BudgetTracker | None = None,
) -> UnitLoopContext:
    """Project the loop-relevant fields off a document state.

    Args:
        state: Document-level agent state.
        budget_tracker: Per-unit tracker LLM calls should be charged to;
            defaults to the document tracker (sequential pipelines).
    """
    return cls(
        ontology_context_mode=state.ontology_context_mode,
        ontology_selection_user_instruction=(
            state.ontology_selection_user_instruction
        ),
        ontology_context_fixed_ontology_id=(
            state.ontology_context_fixed_ontology_id
        ),
        reduced_ontology_artifacts=document_ontology_access(
            state
        ).reduced_artifacts(),
        budget_tracker=budget_tracker
        if budget_tracker is not None
        else state.budget_tracker,
    )
reduced_artifacts()

Reduced document ontology artifacts (shared, read-only).

Source code in ontocast/stategraph/unit_context.py 
72
73
74
def reduced_artifacts(self) -> list[Ontology]:
    """Reduced document ontology artifacts (shared, read-only)."""
    return list(self.reduced_ontology_artifacts)

Functions: