"""Core data structures for OracleMem. The benchmark treats memory writing as selection over virtual experience-representation items. These dataclasses intentionally stay small and JSON-friendly so exact synthetic instances can be inspected and serialized without framework dependencies. """ from __future__ import annotations from dataclasses import asdict, dataclass, field from typing import Any @dataclass(frozen=True) class EvidenceUnit: unit_id: str kind: str text: str proposition_id: str timestamp: int state: str = "current" metadata: dict[str, Any] = field(default_factory=dict) def to_json(self) -> dict[str, Any]: return asdict(self) @dataclass(frozen=True) class Experience: experience_id: str session_id: str timestamp: int text: str visible_unit_ids: tuple[str, ...] metadata: dict[str, Any] = field(default_factory=dict) def to_json(self) -> dict[str, Any]: return asdict(self) @dataclass(frozen=True) class CandidateMemory: candidate_id: str experience_id: str representation: str text: str cost: int coverage: dict[str, float] generator: str = "oracle" metadata: dict[str, Any] = field(default_factory=dict) def to_json(self) -> dict[str, Any]: return asdict(self) @dataclass(frozen=True) class Query: query_id: str text: str category: str required_unit_ids: tuple[str, ...] answer: str metadata: dict[str, Any] = field(default_factory=dict) def to_json(self) -> dict[str, Any]: return asdict(self) @dataclass(frozen=True) class Instance: instance_id: str seed: int units: tuple[EvidenceUnit, ...] experiences: tuple[Experience, ...] candidates: tuple[CandidateMemory, ...] queries: tuple[Query, ...] metadata: dict[str, Any] = field(default_factory=dict) def to_json(self) -> dict[str, Any]: return asdict(self) @dataclass(frozen=True) class SolverResult: method: str budget: int selected_ids: tuple[str, ...] utility: float cost: int ratio: float | None = None ratio_basis: str = "none" metadata: dict[str, Any] = field(default_factory=dict) def to_json(self) -> dict[str, Any]: return asdict(self)