Datasets:
File size: 11,316 Bytes
e1ced61 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 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 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 | """Crash-resumable, idempotent completion-token accounting."""
from __future__ import annotations
import json
from collections.abc import Iterable
from dataclasses import dataclass
from pathlib import Path
from typing import Any
from ..atomic_io import atomic_write_json
from ..hashing import canonical_json_hash, is_sha256
class LedgerError(RuntimeError):
"""Raised on cap overflow, identity drift, or corrupted resume state."""
@dataclass(frozen=True)
class CompletionEntry:
completion_id: str
token_count: int
slot_id: str | None = None
generation_index: int | None = None
completion_sha256: str | None = None
def to_dict(self) -> dict[str, Any]:
row: dict[str, Any] = {
"completion_id": self.completion_id,
"token_count": self.token_count,
}
if self.slot_id is not None:
row["slot_id"] = self.slot_id
if self.generation_index is not None:
row["generation_index"] = self.generation_index
if self.completion_sha256 is not None:
row["completion_sha256"] = self.completion_sha256
return row
def completion_id(run_id: str, slot_id: str, generation_index: int) -> str:
if not run_id or not slot_id:
raise LedgerError("run_id and slot_id must be non-empty")
if (
isinstance(generation_index, bool)
or not isinstance(generation_index, int)
or generation_index < 0
):
raise LedgerError("generation_index must be a non-negative integer")
return canonical_json_hash(
{
"schema_version": 1,
"run_id": run_id,
"slot_id": slot_id,
"generation_index": generation_index,
}
)
class CompletionTokenLedger:
"""Idempotent ledger whose state can be atomically checkpointed and resumed."""
SCHEMA_VERSION = 1
def __init__(
self,
*,
run_id: str,
max_tokens: int,
config_sha256: str | None = None,
data_manifest_sha256: str | None = None,
comparison_slot_manifest_sha256: str | None = None,
) -> None:
if not run_id:
raise LedgerError("run_id must be non-empty")
if isinstance(max_tokens, bool) or not isinstance(max_tokens, int) or max_tokens <= 0:
raise LedgerError("max_tokens must be a positive integer")
for name, value in (
("config_sha256", config_sha256),
("data_manifest_sha256", data_manifest_sha256),
("comparison_slot_manifest_sha256", comparison_slot_manifest_sha256),
):
if value is not None and not is_sha256(value):
raise LedgerError(f"{name} is not a SHA-256")
self.run_id = run_id
self.max_tokens = max_tokens
self.config_sha256 = config_sha256
self.data_manifest_sha256 = data_manifest_sha256
self.comparison_slot_manifest_sha256 = comparison_slot_manifest_sha256
self._entries: dict[str, CompletionEntry] = {}
self._consumed = 0
@property
def consumed_tokens(self) -> int:
return self._consumed
@property
def remaining_tokens(self) -> int:
return self.max_tokens - self._consumed
@property
def exhausted(self) -> bool:
return self._consumed >= self.max_tokens
@property
def completion_count(self) -> int:
return len(self._entries)
def record(
self,
completion_key: str,
token_count: int,
*,
slot_id: str | None = None,
generation_index: int | None = None,
completion_sha256: str | None = None,
) -> bool:
"""Record once; exact duplicates are resume-safe no-ops.
Returns ``True`` for a new record and ``False`` for an identical replay.
Reusing an ID with different metadata is corruption and fails.
"""
return self.record_many(
[
CompletionEntry(
completion_id=completion_key,
token_count=token_count,
slot_id=slot_id,
generation_index=generation_index,
completion_sha256=completion_sha256,
)
]
)[0]
def record_many(self, entries: Iterable[CompletionEntry]) -> tuple[bool, ...]:
pending = list(entries)
seen_pending: dict[str, CompletionEntry] = {}
results: list[bool] = []
new_tokens = 0
for entry in pending:
if not isinstance(entry.completion_id, str) or not entry.completion_id:
raise LedgerError("completion_id must be non-empty")
if (
isinstance(entry.token_count, bool)
or not isinstance(entry.token_count, int)
or entry.token_count < 0
):
raise LedgerError("token_count must be a non-negative integer")
if entry.generation_index is not None and (
isinstance(entry.generation_index, bool)
or not isinstance(entry.generation_index, int)
or entry.generation_index < 0
):
raise LedgerError("generation_index must be a non-negative integer")
if entry.slot_id is not None and (
not isinstance(entry.slot_id, str) or not entry.slot_id
):
raise LedgerError("slot_id must be a non-empty string when supplied")
if entry.completion_sha256 is not None and not is_sha256(entry.completion_sha256):
raise LedgerError("completion_sha256 must be a SHA-256 when supplied")
existing = self._entries.get(entry.completion_id) or seen_pending.get(
entry.completion_id
)
if existing is not None:
if existing != entry:
raise LedgerError(
f"completion_id {entry.completion_id} was reused with different metadata"
)
results.append(False)
continue
seen_pending[entry.completion_id] = entry
new_tokens += entry.token_count
results.append(True)
if self._consumed + new_tokens > self.max_tokens:
raise LedgerError(
f"completion-token cap exceeded: {self._consumed}+{new_tokens}>{self.max_tokens}"
)
for entry in seen_pending.values():
self._entries[entry.completion_id] = entry
self._consumed += new_tokens
return tuple(results)
def to_dict(self) -> dict[str, Any]:
entries = [self._entries[key].to_dict() for key in sorted(self._entries)]
return {
"schema_version": self.SCHEMA_VERSION,
"run_id": self.run_id,
"max_tokens": self.max_tokens,
"consumed_tokens": self._consumed,
"config_sha256": self.config_sha256,
"data_manifest_sha256": self.data_manifest_sha256,
"comparison_slot_manifest_sha256": self.comparison_slot_manifest_sha256,
"entries": entries,
}
def save(self, path: str | Path) -> None:
atomic_write_json(path, self.to_dict())
@classmethod
def load(
cls,
path: str | Path,
*,
expected_run_id: str | None = None,
expected_max_tokens: int | None = None,
expected_config_sha256: str | None = None,
expected_data_manifest_sha256: str | None = None,
expected_comparison_slot_manifest_sha256: str | None = None,
) -> CompletionTokenLedger:
source = Path(path)
try:
value = json.loads(source.read_text(encoding="utf-8"))
except (OSError, json.JSONDecodeError) as exc:
raise LedgerError(f"cannot load ledger {source}: {exc}") from exc
if not isinstance(value, dict) or value.get("schema_version") != cls.SCHEMA_VERSION:
raise LedgerError("unsupported ledger schema")
try:
raw_max_tokens = value.get("max_tokens", 0)
if isinstance(raw_max_tokens, bool) or not isinstance(raw_max_tokens, int):
raise TypeError("max_tokens is not an integer")
ledger = cls(
run_id=str(value.get("run_id", "")),
max_tokens=raw_max_tokens,
config_sha256=value.get("config_sha256"),
data_manifest_sha256=value.get("data_manifest_sha256"),
comparison_slot_manifest_sha256=value.get("comparison_slot_manifest_sha256"),
)
except (TypeError, ValueError, LedgerError) as exc:
raise LedgerError(f"invalid ledger header: {exc}") from exc
checks = (
("run_id", expected_run_id, ledger.run_id),
("max_tokens", expected_max_tokens, ledger.max_tokens),
("config_sha256", expected_config_sha256, ledger.config_sha256),
(
"data_manifest_sha256",
expected_data_manifest_sha256,
ledger.data_manifest_sha256,
),
(
"comparison_slot_manifest_sha256",
expected_comparison_slot_manifest_sha256,
ledger.comparison_slot_manifest_sha256,
),
)
for name, expected, actual in checks:
if expected is not None and expected != actual:
raise LedgerError(f"resume {name} mismatch: {actual!r} != {expected!r}")
raw_entries = value.get("entries")
if not isinstance(raw_entries, list):
raise LedgerError("ledger entries must be a list")
entries: list[CompletionEntry] = []
for raw in raw_entries:
if not isinstance(raw, dict):
raise LedgerError("ledger entry must be an object")
try:
token_count = raw.get("token_count", -1)
generation_index = raw.get("generation_index")
if isinstance(token_count, bool) or not isinstance(token_count, int):
raise TypeError("token_count is not an integer")
if generation_index is not None and (
isinstance(generation_index, bool) or not isinstance(generation_index, int)
):
raise TypeError("generation_index is not an integer")
entries.append(
CompletionEntry(
completion_id=str(raw.get("completion_id", "")),
token_count=token_count,
slot_id=raw.get("slot_id"),
generation_index=generation_index,
completion_sha256=raw.get("completion_sha256"),
)
)
except (TypeError, ValueError) as exc:
raise LedgerError(f"invalid ledger entry: {exc}") from exc
ledger.record_many(entries)
raw_consumed = value.get("consumed_tokens")
if isinstance(raw_consumed, bool) or not isinstance(raw_consumed, int):
raise LedgerError("ledger consumed_tokens must be an integer")
if ledger.consumed_tokens != raw_consumed:
raise LedgerError("ledger consumed_tokens does not equal entry sum")
return ledger
|