wushenghua/sub2api
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
4,022 Commits 1 Branch 0 Tags
98feeccbe132bb85997b3c028c4eec30a9b9aae0
Commit Graph
2960 Commits
Author SHA1 Message Date
lyen1688 8a7269f539 fix: sanitize verbose OpenAI response failed events 2026-06-26 12:29:58 +08:00
Wesley LiddickandGitHub 683a8d8096 Merge pull request #3421 from syx0310/fork/openai-codex-pat-auth-upstream 
feat: add codex personal access token auth
 2026-06-26 11:15:38 +08:00
Wesley LiddickandGitHub 54ef446c1d Merge pull request #3469 from wucm667/fix/responses-passthrough-duplicate-function-args 
fix(openai): Responses passthrough 不再重复发出 function_call arguments，修复 Codex CLI 解析失败
 2026-06-26 10:58:36 +08:00
Wesley LiddickandGitHub fd9c945fab Merge pull request #3470 from wucm667/fix/order-currency-symbol-from-field 
fix(admin): 订单金额币种符号读取 currency 字段，修复 USD 渠道显示为 ¥
 2026-06-26 09:46:02 +08:00
Wesley LiddickandGitHub 94c059c99e Merge pull request #3471 from wucm667/fix/antigravity-standard-tier-empty-project-fallback 
fix(antigravity): standard-tier 账号 project_id 为空时兜底，避免全部请求 500
 2026-06-26 09:45:54 +08:00
Wesley LiddickandGitHub 1721b2b2d1 Merge pull request #3404 from wucm667/fix/balance-preflight-prevent-overdraft 
fix(billing): 余额扣减增加充足性校验，防止余额透支为负
 2026-06-26 09:45:46 +08:00
Wesley LiddickandGitHub 32ddd5c60b Merge pull request #3405 from wucm667/fix/subscription-order-apply-recharge-multiplier 
fix(payment): 订阅套餐订单应用 BALANCE_RECHARGE_MULTIPLIER 汇率换算
 2026-06-26 09:45:38 +08:00
Wesley LiddickandGitHub d53dd112ce Merge pull request #3442 from hsn8086/fix/openai-chat-transport-failover 
fix(openai): fail over on chat completions transport errors
 2026-06-26 09:43:00 +08:00
Wesley LiddickandGitHub 36a352f5eb Merge pull request #3429 from 404QAQ/fix/images-content-refusal-passthrough 
fix(images): pass content-moderation refusals through as 400 instead of retrying
 2026-06-26 09:42:50 +08:00
Wesley LiddickandGitHub 9983f5b80a Merge pull request #3428 from Milesians/codex/fix-subscription-plan-validity-unit 
fix(payment): support plural subscription validity units
 2026-06-26 09:42:42 +08:00
Wesley LiddickandGitHub 6db9e42352 Merge pull request #3476 from zeyugao/main 
Detect OpenAI overloaded errors by structured code
 2026-06-26 09:42:22 +08:00
Wesley LiddickandGitHub fd0da2570d Merge pull request #3481 from visa2/fix/gateway-glm-codex-tool-args-doubled 
fix(apicompat): avoid doubling tool_call arguments from single-chunk upstreams (Codex + GLM)
 2026-06-26 09:42:13 +08:00
Wesley LiddickandGitHub 50c3e52799 Merge pull request #3460 from StarryKira/codex/fix-responses-cache-input 
[codex] fix responses cache input for chat completions bridge
 2026-06-26 09:42:04 +08:00
Wesley LiddickandGitHub db1813f7a3 Merge pull request #3434 from feitianbubu/fix/codex-cli-only-chat-completions 
fix(gateway): enforce codex_cli_only restriction on /v1/chat/completions
 2026-06-26 09:41:46 +08:00
Wesley LiddickandGitHub 6d50934de6 Merge pull request #3457 from feitianbubu/fix/admin-usage-cache-breakdown 
fix(admin/usage): populate cache creation/read token breakdown in stats
 2026-06-26 09:41:36 +08:00
Wesley LiddickandGitHub 27278f62ed Merge pull request #3467 from jianjianai/fix/invalid-refresh-token-nonretryable 
fix(token-refresh): treat refresh_token_invalidated as non-retryable / 添加 refresh_token_invalidated 到不可重试列表
 2026-06-26 09:40:04 +08:00
Wesley LiddickandGitHub 38355ed204 Merge pull request #3436 from feitianbubu/fix/email-identity-error-shadowing 
fix(auth): stop swallowing email auth identity create error via shadowed err
 2026-06-26 09:38:49 +08:00
shaw c9f42e1f77 fix(lint): gofmt token_refresh_service.go after refresh_token_invalidated addition 2026-06-26 09:34:20 +08:00
visa2andClaude Opus 4.8 29122e3051 fix(apicompat): avoid doubling tool_call arguments from single-chunk upstreams 
When converting a Chat Completions stream into Responses events, the first
tool_call delta chunk was copied wholesale into stream state (including
function.arguments), then the same chunk's arguments were accumulated again by
the shared `+=` block. For OpenAI this is harmless because its first tool_call
chunk carries empty arguments, but upstreams that pack id+name+arguments into a
single chunk (e.g. GLM/Zhipu) end up with doubled arguments such as
{"cmd":"ls"}{"cmd":"ls"}. Codex then fails to parse the tool call with
"trailing characters", breaking every tool invocation.

Reset the copied arguments so the shared accumulator counts them exactly once,
keeping the emitted delta and the final done/arguments consistent.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
 2026-06-26 00:58:38 +08:00
Elsa cc7612bdbd Detect OpenAI overloaded error codes 2026-06-25 21:18:39 +08:00
DaydreamCodingandClaude Opus 4.8 0112782049 fix(gateway-openai): codex spark 剥离 image_generation 工具，修复 502 
gpt-5.3-codex-spark 为 text-only，不支持 image_generation 工具；codex CLI 默认
在 /responses 的 tools[] 携带该工具，导致 ChatGPT 后端返回 400
invalid_request_error(param=tools)，被 mapUpstreamError 兜底成 502
"Upstream request failed"（真实 400 见 ops_error_logs.upstream_errors）。

新增 stripCodexSparkImageGenerationTools，对 spark 剥离客户端携带的
image_generation 工具（tools 清空则删键），接入三条路径：
- applyCodexOAuthTransformWithOptions：OAuth 全入口(responses/messages/chat-completions)
- buildUpstreamRequestOpenAIPassthrough 无条件 spark 块：覆盖 OAuth+APIKey
  /responses，不受 image-generation 开关影响
- WS stripCodexSparkImageGenerationToolFromRawPayload：覆盖 WebSocket /responses

bridge 注入(ensureOpenAIResponsesImageGenerationTool)本就跳过 spark；真正的缺口
在 normalizeOpenAIResponsesImageGenerationTools 会规范化并保留客户端工具。
临时绕过=换非 spark 模型；关 bridge/allow_image_generation 无效（工具是客户端带的）。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-25 17:27:19 +08:00
wucm667 650c50e34b fix(antigravity): add project fallback for standard tier 2026-06-25 16:29:33 +08:00
wucm667 55242ffac1 fix(admin): 订单金额币种符号读取 currency 字段 2026-06-25 16:23:45 +08:00
wucm667 2b49d662cc fix(openai): dedupe passthrough function call args 2026-06-25 16:18:11 +08:00
简简aw 0a97a5f461 fix(token-refresh): treat refresh_token_invalidated as non-retryable 2026-06-25 15:12:21 +08:00
shaw 00d68ff6df feat(openai): add GPT-5.5 codex instructions and use as latest fallback 2026-06-25 11:44:42 +08:00
haruka dbdbfb1122 fix: avoid default codex instructions for chat bridge 2026-06-25 02:32:15 +08:00
feitianbubu 4567f6582b test(admin/usage): update sqlmock rows for cache breakdown columns 2026-06-24 23:59:22 +08:00
feitianbubu 063454ae96 fix(admin/usage): populate cache creation/read token breakdown in stats 2026-06-24 23:43:03 +08:00
hsn 65fa728921 fix(openai): fail over on chat transport errors 2026-06-23 22:53:20 +08:00
feitianbubu 82576e0a38 fix(auth): stop swallowing email auth identity create error via shadowed err 2026-06-23 19:29:17 +08:00
feitianbubu ae5e980dd1 fix(gateway): enforce codex_cli_only restriction on /v1/chat/completions 2026-06-23 18:38:52 +08:00
404QAQ 9491de0a39 fix(images): pass content-moderation refusals through as 400 instead of retrying 
When gpt-image upstream returns response.completed with no image but a text
refusal (content moderation, e.g. the model replies "this request was judged
unsafe to generate"), the soft-failure path treated it as a probabilistic
upstream failure and returned a retryable UpstreamFailoverError (502). Retrying
or switching accounts is futile for a content-policy block — it just burns
other accounts' quota and still surfaces an opaque 502 to the client.

Distinguish the two no-image cases:
  (A) model text refusal  -> 400 content_policy_violation, no retry, refusal
      reason passed through to the client.
  (B) truly empty response -> unchanged: retryable UpstreamFailoverError (502).

Adds extractOpenAIImagesModelRefusal (extracts the refusal text from
output_text.delta / message output_text, capped at 600 chars) and two unit
tests; the existing empty-response retry test is unaffected.
 2026-06-23 16:57:16 +08:00
milesians 147c1879d9 fix(payment): support plural subscription validity units 2026-06-23 15:22:55 +08:00
syx0310 32df33a1c3 feat: add codex personal access token auth 2026-06-22 16:07:41 +00:00
wucm667 c6f375d3ab fix(payment): 订阅订单应用充值汇率换算 2026-06-22 10:33:26 +08:00
wucm667 9f5b57fc96 fix(billing): 防止余额计费持续透支 2026-06-22 10:30:27 +08:00
github-actions[bot] d430343f51 chore: sync VERSION to 0.1.138 [skip ci] 2026-06-22 01:24:13 +00:00
shaw 6936687870 fix(lint): check WriteString return in summarizeOpenAIImagesNoOutputBody 
修复 PR #3381 引入的 errcheck lint 错误，satisfy golangci-lint
对 strings.Builder.WriteString 返回值的检查要求。
 2026-06-21 21:43:58 +08:00
Wesley LiddickandGitHub f079742bab Merge pull request #3371 from cugxuan/feat/subscription-affiliate-rebate 
feat: apply affiliate rebate to subscription payments
 2026-06-21 21:18:50 +08:00
Wesley LiddickandGitHub a560d2f91e Merge pull request #3363 from kangjwme/feat/prefer-soonest-reset-scheduling 
feat(scheduling): opt-in "prefer soonest reset" account selection
 2026-06-21 21:17:44 +08:00
Wesley LiddickandGitHub 6e27b82335 Merge pull request #3381 from 404QAQ/fix/images-incomplete-failover 
fix(images): 识别 response.incomplete 触发 failover + 记录软失败上游响应
 2026-06-21 21:14:48 +08:00
Wesley LiddickandGitHub f6d734ec23 Merge pull request #3308 from a0yark/fix/gemini-tool-schema-cleanup 
fix(gemini): clean unsupported tool schema fields
 2026-06-21 21:14:09 +08:00
Wesley LiddickandGitHub af1a032d35 Merge pull request #3375 from StarryKira/fix/3358-vertex-beta-and-cch 
fix(gateway): filter anthropic-beta on the Vertex Anthropic path + drop cch sign (#3358)
 2026-06-21 21:13:14 +08:00
Wesley LiddickandGitHub 93de19665b Merge pull request #3359 from alfadb/fix/glm-effort-mapping 
修复 GLM 推理强度映射
 2026-06-21 21:12:46 +08:00
Wesley LiddickandGitHub bed12b7016 Merge pull request #3360 from feitianbubu/fix/auto-mode-cc-entrypoint-ide 
fix(gateway): in auto mode recognize Claude Code IDE clients via any cc_entrypoint
 2026-06-21 21:12:33 +08:00
Wesley LiddickandGitHub f597e926da Merge pull request #3335 from FjlI5/fix/openai-upstream-endpoint-logging 
fix: 修正 chat-only API-key 账号上游端点被误记为 /v1/responses
 2026-06-21 21:12:15 +08:00
Wesley LiddickandGitHub 7b5fbe5197 Merge pull request #3364 from feitianbubu/fix/promo-clear-expiry 
fix(promo): allow clearing promo code expiry on edit
 2026-06-21 21:12:02 +08:00
404QAQ b0d5592ae2 fix(images): 识别 response.incomplete + 记录软失败上游响应 
修复 gpt-image-2 大图/编辑请求 502 报错（社区 issue #2232/#3135/#2516，
现象：upstream did not return image output）。根因是上游生成超时/截断时返回
response.incomplete，旧逻辑只认 error/response.failed，导致：
1) 软失败报成模糊 502 且不触发 failover 换账号重试
2) 上游真实响应未记录，ops_error_logs 里上游信息全空，无法排查

- openAIImagesUpstreamErrorFromSSEPayload 增加 response.incomplete 识别：
  生成超时/截断(max_output_tokens 等) → 可重试 502 触发 failover；
  content_filter/moderation → 400 不重试
- 软失败兜底(无图无标准错误)记录上游诊断摘要到 ops（last_event/status/
  incomplete_reason/body 片段），非流式+流式两条路径一致
- summarizeOpenAIImagesNoOutputBody 提取诊断信息，body 截断上限 1KB

基于官方 v0.1.137 净分支。测试: 5 个新单测 + 图片回归通过 (-tags unit)
 2026-06-21 08:34:12 +08:00
harukaandClaude Opus 4.8 5cb8cdd36c test(claude-code): detection recognizes the new-CLI billing block (no cch) 
Locks in that Claude Code detection keys on the billing block prefix +
cc_entrypoint=cli, not on the cch field that the new CLI (and now our own
mimicry) no longer sends:

- BillingBlockRecognizedWithoutCCH: an identity-prose-less sub-request whose
  system block is `x-anthropic-billing-header: cc_version=...; cc_entrypoint=cli;`
  (no cch) is still detected as Claude Code.
- NoCCHBlockStillRequiresClaudeCodeUA: dropping cch did not loosen detection —
  a non-claude-cli UA is still rejected, so ClaudeCodeOnly groups can't be spoofed.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
 2026-06-19 08:02:44 -07:00