src/pqc_rag_signing/chunk.py
| 1 | """Signed chunk data structures.""" |
| 2 | |
| 3 | from __future__ import annotations |
| 4 | |
| 5 | import hashlib |
| 6 | import json |
| 7 | from dataclasses import asdict, dataclass, field |
| 8 | from typing import Any |
| 9 | |
| 10 | |
| 11 | @dataclass(frozen=True) |
| 12 | class ChunkMetadata: |
| 13 | """Metadata about a chunk's source (document, position, etc.).""" |
| 14 | |
| 15 | source: str |
| 16 | chunk_index: int |
| 17 | total_chunks: int |
| 18 | start_offset: int = 0 |
| 19 | end_offset: int = 0 |
| 20 | extra: dict[str, Any] = field(default_factory=dict) |
| 21 | |
| 22 | def to_dict(self) -> dict[str, Any]: |
| 23 | return asdict(self) |
| 24 | |
| 25 | |
| 26 | @dataclass |
| 27 | class SignedChunk: |
| 28 | """A chunk of text with a PQC signature envelope. |
| 29 | |
| 30 | This is the unit stored in a vector DB. The ``text`` is the content used |
| 31 | for embedding; the ``_pqc`` envelope is verified at retrieval time. |
| 32 | """ |
| 33 | |
| 34 | chunk_id: str |
| 35 | text: str |
| 36 | metadata: ChunkMetadata |
| 37 | content_hash: str |
| 38 | signer_did: str |
| 39 | algorithm: str |
| 40 | signature: str |
| 41 | public_key: str |
| 42 | signed_at: str |
| 43 | corpus_id: str | None = None |
| 44 | nonce: str = "" |
| 45 | |
| 46 | @staticmethod |
| 47 | def compute_content_hash(text: str, metadata: ChunkMetadata, nonce: str) -> str: |
| 48 | """Canonical SHA3-256 of chunk content + metadata + nonce. |
| 49 | |
| 50 | Deterministic: same (text, metadata, nonce) produces the same hash, always. |
| 51 | """ |
| 52 | canonical = json.dumps( |
| 53 | { |
| 54 | "text": text, |
| 55 | "metadata": metadata.to_dict(), |
| 56 | "nonce": nonce, |
| 57 | }, |
| 58 | sort_keys=True, |
| 59 | separators=(",", ":"), |
| 60 | ensure_ascii=False, |
| 61 | ).encode("utf-8") |
| 62 | return hashlib.sha3_256(canonical).hexdigest() |
| 63 | |
| 64 | def to_dict(self) -> dict[str, Any]: |
| 65 | """Serialize for storage in a vector DB (the whole envelope).""" |
| 66 | return { |
| 67 | "chunk_id": self.chunk_id, |
| 68 | "text": self.text, |
| 69 | "metadata": self.metadata.to_dict(), |
| 70 | "content_hash": self.content_hash, |
| 71 | "signer_did": self.signer_did, |
| 72 | "algorithm": self.algorithm, |
| 73 | "signature": self.signature, |
| 74 | "public_key": self.public_key, |
| 75 | "signed_at": self.signed_at, |
| 76 | "corpus_id": self.corpus_id, |
| 77 | "nonce": self.nonce, |
| 78 | } |
| 79 | |
| 80 | @classmethod |
| 81 | def from_dict(cls, data: dict[str, Any]) -> SignedChunk: |
| 82 | """Deserialize from stored dict.""" |
| 83 | meta = data["metadata"] |
| 84 | if isinstance(meta, dict): |
| 85 | meta = ChunkMetadata( |
| 86 | source=meta["source"], |
| 87 | chunk_index=meta["chunk_index"], |
| 88 | total_chunks=meta["total_chunks"], |
| 89 | start_offset=meta.get("start_offset", 0), |
| 90 | end_offset=meta.get("end_offset", 0), |
| 91 | extra=meta.get("extra", {}), |
| 92 | ) |
| 93 | return cls( |
| 94 | chunk_id=data["chunk_id"], |
| 95 | text=data["text"], |
| 96 | metadata=meta, |
| 97 | content_hash=data["content_hash"], |
| 98 | signer_did=data["signer_did"], |
| 99 | algorithm=data["algorithm"], |
| 100 | signature=data["signature"], |
| 101 | public_key=data["public_key"], |
| 102 | signed_at=data["signed_at"], |
| 103 | corpus_id=data.get("corpus_id"), |
| 104 | nonce=data.get("nonce", ""), |
| 105 | ) |
| 106 | |