codex-lb 0.1.3__py3-none-any.whl → 0.1.5__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -5,7 +5,7 @@ from dataclasses import dataclass
5
5
  from typing import Iterable
6
6
 
7
7
  from app.core.balancer.types import UpstreamError
8
- from app.core.utils.retry import parse_retry_after
8
+ from app.core.utils.retry import backoff_seconds, parse_retry_after
9
9
  from app.db.models import AccountStatus
10
10
 
11
11
  PERMANENT_FAILURE_CODES = {
@@ -22,7 +22,7 @@ class AccountState:
22
22
  account_id: str
23
23
  status: AccountStatus
24
24
  used_percent: float | None = None
25
- reset_at: int | None = None
25
+ reset_at: float | None = None
26
26
  last_error_at: float | None = None
27
27
  last_selected_at: float | None = None
28
28
  error_count: int = 0
@@ -97,18 +97,11 @@ def handle_rate_limit(state: AccountState, error: UpstreamError) -> None:
97
97
  state.status = AccountStatus.RATE_LIMITED
98
98
  state.error_count += 1
99
99
  state.last_error_at = time.time()
100
-
101
- reset_at = _extract_reset_at(error)
102
- if reset_at is not None:
103
- state.reset_at = reset_at
104
- return
105
-
106
100
  message = error.get("message")
107
101
  delay = parse_retry_after(message) if message else None
108
- if delay:
109
- state.reset_at = int(time.time() + delay)
110
- else:
111
- state.reset_at = int(time.time() + 300)
102
+ if delay is None:
103
+ delay = backoff_seconds(state.error_count)
104
+ state.reset_at = time.time() + delay
112
105
 
113
106
 
114
107
  def handle_quota_exceeded(state: AccountState, error: UpstreamError) -> None:
app/core/utils/retry.py CHANGED
@@ -1,8 +1,13 @@
1
1
  from __future__ import annotations
2
2
 
3
+ import random
3
4
  import re
4
5
 
5
6
  _RETRY_PATTERN = re.compile(r"(?i)try again in\s*(\d+(?:\.\d+)?)\s*(s|ms|seconds?)")
7
+ _BACKOFF_INITIAL_DELAY_MS = 200
8
+ _BACKOFF_FACTOR = 2.0
9
+ _BACKOFF_JITTER_MIN = 0.9
10
+ _BACKOFF_JITTER_MAX = 1.1
6
11
 
7
12
 
8
13
  def parse_retry_after(message: str) -> float | None:
@@ -14,3 +19,12 @@ def parse_retry_after(message: str) -> float | None:
14
19
  if unit == "ms":
15
20
  return value / 1000
16
21
  return value
22
+
23
+
24
+ def backoff_seconds(attempt: int) -> float:
25
+ if attempt < 1:
26
+ attempt = 1
27
+ exponent = _BACKOFF_FACTOR ** (attempt - 1)
28
+ base_ms = _BACKOFF_INITIAL_DELAY_MS * exponent
29
+ jitter = random.uniform(_BACKOFF_JITTER_MIN, _BACKOFF_JITTER_MAX)
30
+ return (base_ms * jitter) / 1000.0
app/db/session.py CHANGED
@@ -1,5 +1,6 @@
1
1
  from __future__ import annotations
2
2
 
3
+ import asyncio
3
4
  from pathlib import Path
4
5
  from typing import AsyncIterator
5
6
 
@@ -23,16 +24,33 @@ def _ensure_sqlite_dir(url: str) -> None:
23
24
  Path(path).expanduser().parent.mkdir(parents=True, exist_ok=True)
24
25
 
25
26
 
27
+ async def _safe_rollback(session: AsyncSession) -> None:
28
+ if not session.in_transaction():
29
+ return
30
+ try:
31
+ await asyncio.shield(session.rollback())
32
+ except Exception:
33
+ return
34
+
35
+
36
+ async def _safe_close(session: AsyncSession) -> None:
37
+ try:
38
+ await asyncio.shield(session.close())
39
+ except Exception:
40
+ return
41
+
42
+
26
43
  async def get_session() -> AsyncIterator[AsyncSession]:
27
- async with SessionLocal() as session:
28
- try:
29
- yield session
30
- except Exception:
31
- await session.rollback()
32
- raise
33
- finally:
34
- if session.in_transaction():
35
- await session.rollback()
44
+ session = SessionLocal()
45
+ try:
46
+ yield session
47
+ except BaseException:
48
+ await _safe_rollback(session)
49
+ raise
50
+ finally:
51
+ if session.in_transaction():
52
+ await _safe_rollback(session)
53
+ await _safe_close(session)
36
54
 
37
55
 
38
56
  async def init_db() -> None:
app/dependencies.py CHANGED
@@ -7,7 +7,7 @@ from dataclasses import dataclass
7
7
  from fastapi import Depends
8
8
  from sqlalchemy.ext.asyncio import AsyncSession
9
9
 
10
- from app.db.session import SessionLocal, get_session
10
+ from app.db.session import SessionLocal, _safe_close, _safe_rollback, get_session
11
11
  from app.modules.accounts.repository import AccountsRepository
12
12
  from app.modules.accounts.service import AccountsService
13
13
  from app.modules.oauth.service import OauthService
@@ -87,15 +87,16 @@ def get_usage_context(
87
87
 
88
88
  @asynccontextmanager
89
89
  async def _accounts_repo_context() -> AsyncIterator[AccountsRepository]:
90
- async with SessionLocal() as session:
91
- try:
92
- yield AccountsRepository(session)
93
- except Exception:
94
- await session.rollback()
95
- raise
96
- finally:
97
- if session.in_transaction():
98
- await session.rollback()
90
+ session = SessionLocal()
91
+ try:
92
+ yield AccountsRepository(session)
93
+ except BaseException:
94
+ await _safe_rollback(session)
95
+ raise
96
+ finally:
97
+ if session.in_transaction():
98
+ await _safe_rollback(session)
99
+ await _safe_close(session)
99
100
 
100
101
 
101
102
  def get_oauth_context(
@@ -20,7 +20,7 @@ from app.modules.usage.repository import UsageRepository
20
20
 
21
21
  @dataclass
22
22
  class RuntimeState:
23
- reset_at: int | None = None
23
+ reset_at: float | None = None
24
24
  last_error_at: float | None = None
25
25
  last_selected_at: float | None = None
26
26
  error_count: int = 0
@@ -179,10 +179,10 @@ def _apply_secondary_quota(
179
179
  *,
180
180
  status: AccountStatus,
181
181
  primary_used: float | None,
182
- runtime_reset: int | None,
182
+ runtime_reset: float | None,
183
183
  secondary_used: float | None,
184
184
  secondary_reset: int | None,
185
- ) -> tuple[AccountStatus, float | None, int | None]:
185
+ ) -> tuple[AccountStatus, float | None, float | None]:
186
186
  used_percent = primary_used
187
187
  reset_at = runtime_reset
188
188
 
@@ -1,5 +1,6 @@
1
1
  from __future__ import annotations
2
2
 
3
+ import logging
3
4
  import time
4
5
  from datetime import timedelta
5
6
  from typing import AsyncIterator, Iterable, Mapping
@@ -36,6 +37,8 @@ from app.modules.proxy.usage_updater import UsageUpdater
36
37
  from app.modules.request_logs.repository import RequestLogsRepository
37
38
  from app.modules.usage.repository import UsageRepository
38
39
 
40
+ logger = logging.getLogger(__name__)
41
+
39
42
 
40
43
  class ProxyService:
41
44
  def __init__(
@@ -191,6 +194,7 @@ class ProxyService:
191
194
  yield format_sse_event(event)
192
195
  return
193
196
 
197
+ account_id_value = account.id
194
198
  try:
195
199
  account = await self._ensure_fresh(account)
196
200
  async for line in self._stream_once(
@@ -243,7 +247,15 @@ class ProxyService:
243
247
  await self._load_balancer.mark_permanent_failure(account, exc.code)
244
248
  continue
245
249
  except Exception:
246
- await self._load_balancer.record_error(account)
250
+ try:
251
+ await self._load_balancer.record_error(account)
252
+ except Exception:
253
+ logger.warning(
254
+ "Failed to record proxy error account_id=%s request_id=%s",
255
+ account_id_value,
256
+ request_id,
257
+ exc_info=True,
258
+ )
247
259
  if attempt == max_attempts - 1:
248
260
  event = response_failed_event(
249
261
  "upstream_error",
@@ -267,8 +279,9 @@ class ProxyService:
267
279
  request_id: str,
268
280
  allow_retry: bool,
269
281
  ) -> AsyncIterator[str]:
282
+ account_id_value = account.id
270
283
  access_token = self._encryptor.decrypt(account.access_token_encrypted)
271
- account_id = _header_account_id(account.id)
284
+ account_id = _header_account_id(account_id_value)
272
285
  model = payload.model
273
286
  start = time.monotonic()
274
287
  status = "success"
@@ -341,19 +354,27 @@ class ProxyService:
341
354
  reasoning_tokens = (
342
355
  usage.output_tokens_details.reasoning_tokens if usage and usage.output_tokens_details else None
343
356
  )
344
- await self._logs_repo.add_log(
345
- account_id=account.id,
346
- request_id=request_id,
347
- model=model,
348
- input_tokens=input_tokens,
349
- output_tokens=output_tokens,
350
- cached_input_tokens=cached_input_tokens,
351
- reasoning_tokens=reasoning_tokens,
352
- latency_ms=latency_ms,
353
- status=status,
354
- error_code=error_code,
355
- error_message=error_message,
356
- )
357
+ try:
358
+ await self._logs_repo.add_log(
359
+ account_id=account_id_value,
360
+ request_id=request_id,
361
+ model=model,
362
+ input_tokens=input_tokens,
363
+ output_tokens=output_tokens,
364
+ cached_input_tokens=cached_input_tokens,
365
+ reasoning_tokens=reasoning_tokens,
366
+ latency_ms=latency_ms,
367
+ status=status,
368
+ error_code=error_code,
369
+ error_message=error_message,
370
+ )
371
+ except Exception:
372
+ logger.warning(
373
+ "Failed to persist request log account_id=%s request_id=%s",
374
+ account_id_value,
375
+ request_id,
376
+ exc_info=True,
377
+ )
357
378
 
358
379
  async def _refresh_usage(self, accounts: list[Account]) -> None:
359
380
  latest_usage = await self._usage_repo.latest_by_account(window="primary")
@@ -1,5 +1,6 @@
1
1
  from __future__ import annotations
2
2
 
3
+ import asyncio
3
4
  from datetime import datetime
4
5
 
5
6
  from sqlalchemy import and_, select
@@ -49,9 +50,13 @@ class RequestLogsRepository:
49
50
  requested_at=requested_at or utcnow(),
50
51
  )
51
52
  self._session.add(log)
52
- await self._session.commit()
53
- await self._session.refresh(log)
54
- return log
53
+ try:
54
+ await self._session.commit()
55
+ await self._session.refresh(log)
56
+ return log
57
+ except BaseException:
58
+ await _safe_rollback(self._session)
59
+ raise
55
60
 
56
61
  async def list_recent(
57
62
  self,
@@ -84,3 +89,12 @@ class RequestLogsRepository:
84
89
  stmt = stmt.limit(limit)
85
90
  result = await self._session.execute(stmt)
86
91
  return list(result.scalars().all())
92
+
93
+
94
+ async def _safe_rollback(session: AsyncSession) -> None:
95
+ if not session.in_transaction():
96
+ return
97
+ try:
98
+ await asyncio.shield(session.rollback())
99
+ except Exception:
100
+ return
app/static/index.js CHANGED
@@ -569,16 +569,9 @@
569
569
  return acc;
570
570
  }, {});
571
571
 
572
- const mergeUsageIntoAccounts = (
573
- accounts,
574
- primaryUsage,
575
- secondaryUsage,
576
- summary,
577
- ) => {
572
+ const mergeUsageIntoAccounts = (accounts, primaryUsage, secondaryUsage) => {
578
573
  const primaryMap = buildUsageIndex(primaryUsage || []);
579
574
  const secondaryMap = buildUsageIndex(secondaryUsage || []);
580
- const resetAtPrimary = summary?.primaryWindow?.resetAt ?? null;
581
- const resetAtSecondary = summary?.secondaryWindow?.resetAt ?? null;
582
575
  return accounts.map((account) => {
583
576
  const primaryRow = primaryMap[account.id];
584
577
  const secondaryRow = secondaryMap[account.id];
@@ -598,8 +591,8 @@
598
591
  account.usage?.secondaryRemainingPercent ??
599
592
  0,
600
593
  },
601
- resetAtPrimary: resetAtPrimary ?? account.resetAtPrimary ?? null,
602
- resetAtSecondary: resetAtSecondary ?? account.resetAtSecondary ?? null,
594
+ resetAtPrimary: account.resetAtPrimary ?? null,
595
+ resetAtSecondary: account.resetAtSecondary ?? null,
603
596
  };
604
597
  });
605
598
  };
@@ -1191,7 +1184,6 @@
1191
1184
  accountsResult.value,
1192
1185
  primaryUsage,
1193
1186
  secondaryUsage,
1194
- summary,
1195
1187
  );
1196
1188
  this.applyData(
1197
1189
  {
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: codex-lb
3
- Version: 0.1.3
3
+ Version: 0.1.5
4
4
  Summary: Codex load balancer and proxy for ChatGPT accounts with usage dashboard
5
5
  Author-email: Soju06 <qlskssk@gmail.com>
6
6
  Maintainer-email: Soju06 <qlskssk@gmail.com>
@@ -1,6 +1,6 @@
1
1
  app/__init__.py,sha256=X_4Y93UbuflRBRZd65MtjsVxlZW8d9ZUyi0n83Q-Lrg,89
2
2
  app/cli.py,sha256=gkIAkYOT9SbQjUDnVmwhVKZeKjL3YJCMrOjFINwBx54,544
3
- app/dependencies.py,sha256=43XzEFcOJbfw3Wb1O-fGyLemfgu9RmYPrULUUTPst54,3933
3
+ app/dependencies.py,sha256=9VvBKrUctmuab793Yq2MsHmAh7upXt0viTeofdop8lA,3966
4
4
  app/main.py,sha256=JMvTfBdS8wTDNqjQ5QSdqKjjhE0CkpfkbFW6fWNMpy0,4351
5
5
  app/core/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
6
6
  app/core/crypto.py,sha256=zUz2GVqXigzXB5zX2Diq2cR41Kl37bBqxJiZmWGiIcY,1145
@@ -10,7 +10,7 @@ app/core/auth/__init__.py,sha256=wR52dbGd66W1haAOUXn5RMzn3bj-EeXm4vm7zqI3kHE,281
10
10
  app/core/auth/models.py,sha256=iyygu0uHK7fu8txyIyCkXmlgqYPbqEZu3IFysD1t-gk,1557
11
11
  app/core/auth/refresh.py,sha256=jhYxT2mQFX4CCvuBQfYNcI53iOtWa0vSkTuZAfbpd9k,5105
12
12
  app/core/balancer/__init__.py,sha256=bGy7gITRPObt6P9NdG3pg8t3qWoSW0XWoADqfNYrhdg,405
13
- app/core/balancer/logic.py,sha256=JFQjzYFzTFRl2Ps24Me-KFZ8jVLMe3Jj5Xzm3aJzFs4,5168
13
+ app/core/balancer/logic.py,sha256=aBuluReYRaYRmXNBntyUhOy1oPsK6Ol4Yt6WSkBLzn4,5059
14
14
  app/core/balancer/types.py,sha256=gDgjlTy-NH3YhHYl2-YYpIabnckN9Q8-4cRy6S1u0K4,191
15
15
  app/core/clients/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
16
16
  app/core/clients/http.py,sha256=yfFwsaIZNbSmNtyV02WnMKn00JdoNNSlSIx5xB3QY4o,987
@@ -30,12 +30,12 @@ app/core/usage/pricing.py,sha256=6p8rJ26Gk61mz2t_h9sa0T7NiPDUTiNpzoDewMzT6E0,546
30
30
  app/core/usage/types.py,sha256=CbFF6JYSLvALa2P0qYuj9J9b_w53Fgjg0JOu4RzMWbI,2104
31
31
  app/core/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
32
32
  app/core/utils/request_id.py,sha256=gxafI-Se8dRQTir-HNGelMzC9S2gwYZntTkVzEqzp_I,705
33
- app/core/utils/retry.py,sha256=E1s397Bse072sdeEZ2uri5sPAOlTIdyccB-F9jh_E64,408
33
+ app/core/utils/retry.py,sha256=UmBap1Wh-CBT7r4fHzVb_PI9-LR9-HjUtDzRnhRjP2U,822
34
34
  app/core/utils/sse.py,sha256=ER3_7mm50BUC1CPIKoVOmQikz-vlEWKNSRDEVqzI_KA,387
35
35
  app/core/utils/time.py,sha256=B6FfSe43Eq_puE6eourly1X3gajyihK2VOAwJ8M3wyI,497
36
36
  app/db/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
37
37
  app/db/models.py,sha256=JCknQCuzjHfgyuSzjHqBmeIE-0XisIAQGEhEWqmzabs,3841
38
- app/db/session.py,sha256=9UBO4fZKWbN3ihbhNxlkFDDXBMRnlV4M0INvatAeyn0,1211
38
+ app/db/session.py,sha256=te9-diH008J-YUAnmE7-OCBUII4LB2jgLlV_ipgoej4,1584
39
39
  app/modules/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
40
40
  app/modules/accounts/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
41
41
  app/modules/accounts/api.py,sha256=rSkYrg3_p_JE1kHZ4xXa5lBFTdT40v1oNAmlTDGOguY,2807
@@ -52,14 +52,14 @@ app/modules/oauth/templates/oauth_success.html,sha256=YNSGUIozcZEJQjpFtM2sgF4n8j
52
52
  app/modules/proxy/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
53
53
  app/modules/proxy/api.py,sha256=BR_qNlNZg2Ft5GYQ-7AzlLlJFf6RVJmlzJzFr_KHUIc,2921
54
54
  app/modules/proxy/auth_manager.py,sha256=sFcbStd-OK0g8S56z7XcSkFJ22kEirBU0fYdz2drS0o,2185
55
- app/modules/proxy/load_balancer.py,sha256=k53RXn9v9fQPXEJ3zWbkhh_Jd82PK1tUef3ZvCdSzog,7441
55
+ app/modules/proxy/load_balancer.py,sha256=L1a7duA4mcR61huXKU052QNz4JAGgUwhMuOPpv5BV4w,7447
56
56
  app/modules/proxy/schemas.py,sha256=55pXtUCl2R_93kAPOJJ7Ji4Jn3qVu10vq2KSCCkNdp4,2748
57
- app/modules/proxy/service.py,sha256=P5TyqZIo63Yz70LGAB5Vb5Um9Wy7Toj9qEyHJ39Qero,26018
57
+ app/modules/proxy/service.py,sha256=XBpL1OIZuzss8nvsYajC5KsaRdGztjci5RgnWORZ1ss,26816
58
58
  app/modules/proxy/types.py,sha256=iqEyoO8vGr8N5oEzUSvVWCai7UZbJAU62IvO7JNS9qs,927
59
59
  app/modules/proxy/usage_updater.py,sha256=t5wpn3SmPnTXceiQFF0tw-9V5oPPhP4C0anwbI7T5v0,5588
60
60
  app/modules/request_logs/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
61
61
  app/modules/request_logs/api.py,sha256=6nV2Uv_hnK7WI3gNpKrgTx4MyUQIXk1QxKp40nij0Xo,1037
62
- app/modules/request_logs/repository.py,sha256=S86lCDNVAb4N4SvZs5Cnj96ypGzV39r3syS649iOoQ8,2991
62
+ app/modules/request_logs/repository.py,sha256=SnwDwD84wgslmdIgYFGD8ReUYmIyhxT6MOjFAolLBC8,3333
63
63
  app/modules/request_logs/schemas.py,sha256=GSCi4TEWMmQ-THsQl2irRrA_msU8jzsqKSBDFn7hiJU,592
64
64
  app/modules/request_logs/service.py,sha256=SuJeeFJy0qhMe6cUgQ3-zPrENoPlcNnDmTvG7aYsLmE,2386
65
65
  app/modules/shared/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
@@ -72,9 +72,9 @@ app/modules/usage/service.py,sha256=_LQA3-L1FxIpzPoIQ-rxz3n3KjiHpTSxRimZapzwx3o,
72
72
  app/static/7.css,sha256=9EHW2Ouff2fRXgcQbYuCuglxTFgQZGNLLTXKTS6S5aI,80687
73
73
  app/static/index.css,sha256=ct1FfBg0PY7c6KxSS6A7JM_WDwdeJqhQa4pXTgyJ19Q,8786
74
74
  app/static/index.html,sha256=FmWQ0l40rTBsHIR9i0ttGTzG17ePASsO2VDv6Q48CNw,24477
75
- app/static/index.js,sha256=L6Gie0iSAHyrPLvO_QtzEis4rxhVqNe2m-g4h3lB6GU,52419
76
- codex_lb-0.1.3.dist-info/METADATA,sha256=f7L6g4cETQ9XBxXs-pzJKK1hdnuF9FKIhe6kZTLDnM8,3828
77
- codex_lb-0.1.3.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
78
- codex_lb-0.1.3.dist-info/entry_points.txt,sha256=SEa5T6Uz2Fhy574No6Y0XyGmYi3PXLrhu2xStJTqyI8,42
79
- codex_lb-0.1.3.dist-info/licenses/LICENSE,sha256=cHPibxiL0TXwrUX_kNY6ym544EX1UCzKhxdaca5cFuk,1062
80
- codex_lb-0.1.3.dist-info/RECORD,,
75
+ app/static/index.js,sha256=2gbVWp_Ot00NLdPpLE2Cc-YWk38H04x8TZw3OFO2SKk,52209
76
+ codex_lb-0.1.5.dist-info/METADATA,sha256=Cjw3NQBgM5hFgPAQUZ81PFBOI2mAq8SiwViALHOkxxA,3828
77
+ codex_lb-0.1.5.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
78
+ codex_lb-0.1.5.dist-info/entry_points.txt,sha256=SEa5T6Uz2Fhy574No6Y0XyGmYi3PXLrhu2xStJTqyI8,42
79
+ codex_lb-0.1.5.dist-info/licenses/LICENSE,sha256=cHPibxiL0TXwrUX_kNY6ym544EX1UCzKhxdaca5cFuk,1062
80
+ codex_lb-0.1.5.dist-info/RECORD,,