@@ -0,0 +1,76 @@ |
| 1 | +"""Minimal section contract for attribution probes. |
| 2 | + |
| 3 | +The flagship B1 ``section_internalization`` probe needs *structured* |
| 4 | +input — a section has an id, a kind, content text, and possibly some |
| 5 | +Q/A pairs or chosen/rejected triples. sway defines this shape here so |
| 6 | +the probes stay oblivious to the upstream (``.dlm`` parser, custom |
| 7 | +loaders, synthetic test fixtures). |
| 8 | + |
| 9 | +Field names are aligned with :mod:`dlm.doc.sections` but this module |
| 10 | +does not import ``dlm`` — the bridge at |
| 11 | +:mod:`dlm_sway.integrations.dlm` does the adaptation. |
| 12 | +""" |
| 13 | + |
| 14 | +from __future__ import annotations |
| 15 | + |
| 16 | +from dataclasses import dataclass, field |
| 17 | +from typing import Literal |
| 18 | + |
| 19 | +SectionKind = Literal["prose", "instruction", "preference"] |
| 20 | + |
| 21 | + |
| 22 | +@dataclass(frozen=True, slots=True) |
| 23 | +class SectionProbe: |
| 24 | + """A ``(prompt, gold)`` pair lifted from an INSTRUCTION section.""" |
| 25 | + |
| 26 | + prompt: str |
| 27 | + gold: str |
| 28 | + |
| 29 | + |
| 30 | +@dataclass(frozen=True, slots=True) |
| 31 | +class SectionPreference: |
| 32 | + """A ``(prompt, chosen, rejected)`` triple from a PREFERENCE section.""" |
| 33 | + |
| 34 | + prompt: str |
| 35 | + chosen: str |
| 36 | + rejected: str |
| 37 | + |
| 38 | + |
| 39 | +@dataclass(frozen=True, slots=True) |
| 40 | +class Section: |
| 41 | + """One typed chunk of a training document. |
| 42 | + |
| 43 | + Attributes |
| 44 | + ---------- |
| 45 | + id: |
| 46 | + Content-addressed identifier. ``.dlm`` uses a 16-hex-char |
| 47 | + sha256 prefix; sway doesn't enforce a format. |
| 48 | + kind: |
| 49 | + Discriminator for which of :attr:`probes` / |
| 50 | + :attr:`preferences` / :attr:`content` is the primary signal. |
| 51 | + content: |
| 52 | + Raw section text. Always populated; used by the rolling-PPL |
| 53 | + path for PROSE sections. |
| 54 | + probes: |
| 55 | + For INSTRUCTION: parsed Q/A pairs. Empty tuple for others. |
| 56 | + preferences: |
| 57 | + For PREFERENCE: parsed chosen/rejected triples. Empty otherwise. |
| 58 | + tag: |
| 59 | + Optional free-form label for the section (e.g., "intro", |
| 60 | + "api-reference"). Surfaces in per-section reports. |
| 61 | + """ |
| 62 | + |
| 63 | + id: str |
| 64 | + kind: SectionKind |
| 65 | + content: str |
| 66 | + probes: tuple[SectionProbe, ...] = field(default_factory=tuple) |
| 67 | + preferences: tuple[SectionPreference, ...] = field(default_factory=tuple) |
| 68 | + tag: str | None = None |
| 69 | + |
| 70 | + |
| 71 | +def filter_kinds( |
| 72 | + sections: tuple[Section, ...], kinds: tuple[SectionKind, ...] |
| 73 | +) -> tuple[Section, ...]: |
| 74 | + """Return only sections whose ``kind`` matches one of ``kinds``.""" |
| 75 | + allow = set(kinds) |
| 76 | + return tuple(s for s in sections if s.kind in allow) |
