Commit Graph
16 Commits
Author SHA1 Message Date
erio 1e2e8b1d6b fix(billing): channel pricing fully overrides image output token price
When channel pricing is configured, image output token price now
correctly overrides all other sources (LiteLLM, group config).
Add ImageOutputPriceExplicit flag to distinguish intentional zero
price from missing price, preventing fallback to output price.

Affects Gemini, OpenAI inline image gen, and Images API paths.
2026-05-31 22:57:56 +08:00
PluviobyteandCursor ed2aac25a6 fix(billing): apply long-context multiplier to cache_creation price
Follow-up to #2816 (already merged): the same long-context pricing
exemption that affected cache_read also applies to all three
cache_creation price fields (standard, 5m ephemeral, 1h ephemeral).
computeCacheCreationCost reads these prices directly from pricing and
never sees the LongContextInputMultiplier that computeTokenBreakdown
applies to inputPrice / outputPrice / cacheReadPrice.

For GPT-5.4 / 5.5 above the 272k threshold, this causes the cache_write
portion of long sessions to be billed at roughly half what it should
be (default multiplier 2.0). Cache writes are conceptually input-side
operations and should share the same long-context treatment as input /
cache_read.

This patch threads an explicit multiplier into computeCacheCreationCost
so the function can be unit-tested in isolation and matches the existing
pattern used for cache_read. computeTokenBreakdown captures the long
context decision once and passes LongContextInputMultiplier when it
applies, 1.0 otherwise.

Adds three regression tests mirroring the #2816 cache_read tests:
- positive: long-context triggered -> cache_creation scaled by 2.0x
- negative: below threshold -> cache_creation stays at base price
- breakdown: 5m + 1h ephemeral prices both scaled when applicable

Refs #2816

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-27 09:59:58 +00:00
SlientRainyDayandCursor b9509e823a fix(billing): apply long-context multiplier to cache_read price
When session long-context pricing is triggered in computeTokenBreakdown
(e.g. GPT-5.4 / GPT-5.5 above the 272k token threshold), the multiplier
was only being applied to InputPricePerToken and OutputPricePerToken.
The cache_read price was left at its base value, so CacheReadCost was
silently undercharged whenever a long-context session also had cache
hits — which is essentially every long Codex / Claude Code session.

Concretely for gpt-5.4 with 300k cache_read tokens, the bug
under-billed the request by exactly 1x the LongContextInputMultiplier
on the cache portion (e.g. 0.075 instead of 0.150 in the regression
test).

Cache reads are conceptually input-side replays, so they should scale
with LongContextInputMultiplier, matching the treatment of
InputPricePerToken.

Adds two regression tests:
- positive: long-context triggered -> cache_read scaled by 2.0x
- negative: below threshold -> cache_read stays at base price

Fixes #2293

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-27 07:09:28 +00:00
2ue 6faa344916 feat: add OpenAI image generation controls 2026-05-05 03:26:54 +08:00
erio bbc4aed3d9 fix(openai): 移除已下线 Codex 模型并修复归一化兜底副作用
- backend: 删除 gpt-5 / 5.1 / 5.1-codex / 5.1-codex-max / 5.1-codex-mini / 5.2-codex / 5.4-nano 的内置映射与 DefaultModels 条目
- backend: normalizeCodexModel 默认兜底由 gpt-5.1 改为 gpt-5.4,gpt-5.3-codex-spark 独立保留映射
- backend: 修复 isOpenAIGPT54Model 与 shouldAutoInjectPromptCacheKeyForCompat 对 claude / gpt-4o 的误判(之前依赖 gpt-5.1 作为非 GPT 族的隐式 sentinel,改后需要显式前缀守卫)
- backend: 清理 billing_service 中已不可达的 fallback 价格与 switch 分支
- frontend: 从白名单、OpenCode 配置、预设映射中移除已下线模型
- 同步更新所有相关单测

Refs: #1758, parallels upstream #1759 but adds downstream guard fixes
2026-04-20 22:01:41 +08:00
erio df57d2776b fix(billing): reject rate_multiplier <= 0 on save; clamp negatives to 0 in compute
分组倍率和用户专属倍率在保存时没有校验,0 会触发计费层的 `<=0 → 1.0`
防御条款,结果订阅/余额分组按标准价扣费;完全是沉默地绕过了业务规则。

- 保存校验(admin_service):CreateGroup / UpdateGroup / BatchSetGroupRateMultipliers /
  UpdateUser.SyncUserGroupRates 全部要求 > 0
- 计算层(billing_service):三处 `<=0 → 1.0` 改为 `<0 → 0`;负数按 0 结算,
  避免配置异常被静默按 1x 收费
- 前端:分组倍率 / 用户专属倍率输入 min 统一到 0.001
- 删除未使用的 IsFreeSubscription 方法

测试:新增 billing_service_rate_multiplier_test.go 端到端验证;更新原有锁定
旧 `<=0 → 1.0` 行为的测试。
2026-04-17 22:06:32 +08:00
erio 9028d2085f test: add unit tests for billing, websearch, and notify systems
Billing (25 tests):
- CalculateCostUnified: nil resolver fallback, token/per_request/image modes
- GetModelPricingWithChannel: nil/partial/full channel overrides
- resolveAccountStatsCost: four-level priority chain integration tests

WebSearch (18 tests):
- PopulateWebSearchUsage: nil input, manager states, QuotaLimit nil/*int64
- ResetWebSearchUsage: nil manager error
- Manager.ResetUsage: nil Redis
- shouldEmulateWebSearch: full decision chain (8 scenarios)

Notify (36 tests):
- ParseNotifyEmails/MarshalNotifyEmails: old/new format, roundtrip
- crossedDownward: boundary values, threshold semantics
- checkQuotaDimCrossings: mixed dimensions, disabled/zero skip
2026-04-14 09:36:40 +08:00
erio 3d4d960d60 fix: gofmt formatting after merge 2026-04-14 09:15:49 +08:00
erio 62e80c602d revert: completely remove all Sora functionality 2026-04-05 17:11:01 +08:00
Remx 42d73118fd feat(openai): 增加 gpt-5.4-mini/nano 模型支持与定价配置
- 接入 gpt-5.4-mini/nano 模型识别与规范化,补充默认模型列表
- 增加 gpt-5.4-mini/nano 输入/缓存命中/输出价格与计费兜底逻辑
- 同步前端模型白名单与 OpenCode 配置
- 补充 service tier(priority/flex) 计费回归测试
2026-03-19 19:03:13 +08:00
yangjianboandClaude Opus 4.6 87f4ed591e fix(billing): 修复 OpenAI fast 档位计费并补齐展示
- 打通 service_tier 在 OpenAI HTTP、WS、passthrough 与 usage 记录中的传递
- 修正 priority/flex 计费逻辑,并将 fast 归一化为 priority
- 在用户端和管理端补齐服务档位与计费明细展示
- 补齐前后端测试,并修复 WS 限流信号重复持久化导致的全量回归失败

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-09 09:51:26 +08:00
yangjianbo f366026435 fix(openai): 修复 gpt-5.4 长上下文计费与快照白名单
补齐 gpt-5.4 fallback 的长上下文计费元信息,\n确保超过 272000 输入 token 时对整次会话应用\n2x 输入与 1.5x 输出计费规则。\n\n同时将官方快照 gpt-5.4-2026-03-05 加入前端\n白名单候选与回归测试,避免 whitelist 模式误拦截。\n\nCo-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

(cherry picked from commit d95497af87f608c6dadcbe7d6e851de9413ae147)
2026-03-06 10:16:23 +08:00
yangjianbo 1a0d4ed668 feat(openai): 增加 gpt-5.4 模型支持与定价配置
- 接入 gpt-5.4 模型识别与规范化,补充默认模型列表
- 增加 gpt-5.4 输入/缓存命中/输出价格与计费兜底逻辑
- 同步前端模型白名单与 OpenCode 上下文窗口(1050000/128000)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
(cherry picked from commit 924476dcac6181cd0f3ee731ec7b73672ff03793)
2026-03-06 10:16:23 +08:00
yangjianbo 48dc011b2a test(admin,service): 修复代理质量与计费单测口径 2026-02-19 21:39:31 +08:00
yangjianboandClaude Opus 4.6 e489996713 test(backend): 补充改动代码单元测试覆盖率至 85%+
新增 48 个测试用例覆盖修复代码的各分支路径:
- subscription_maintenance_queue: nil receiver/task、Stop 幂等、零值参数 (+6)
- billing_service: CalculateCostWithConfig、错误传播、SoraImageCost 等 (+12)
- timing_wheel_service: Schedule/ScheduleRecurring after Stop (+3)
- sora_media_cleanup_service: nil guard、Start/Stop 各分支、timezone (+10)
- sora_gateway_service: normalizeSoraMediaURLs、buildSoraContent 等辅助函数 (+17)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-10 17:52:10 +08:00
yangjianboandClaude Opus 4.6 bb5a5dd65e test: 完善自动化测试体系(7个模块,73个任务)
系统性地修复、补充和强化项目的自动化测试能力:

1. 测试基础设施修复
   - 修复 stubConcurrencyCache 缺失方法和构造函数参数不匹配
   - 创建 testutil 共享包(stubs.go, fixtures.go, httptest.go)
   - 为所有 Stub 添加编译期接口断言

2. 中间件测试补充
   - 新增 JWT 认证中间件测试(有效/过期/篡改/缺失 Token)
   - 补充 rate_limiter 和 recovery 中间件测试场景

3. 网关核心路径测试
   - 新增账户选择、等待队列、流式响应、并发控制、计费、Claude Code 检测测试
   - 覆盖负载均衡、粘性会话、SSE 转发、槽位管理等关键逻辑

4. 前端测试体系(11个新测试文件,163个测试用例)
   - Pinia stores: auth, app, subscriptions
   - API client: 请求拦截器、响应拦截器、401 刷新
   - Router guards: 认证重定向、管理员权限、简易模式限制
   - Composables: useForm, useTableLoader, useClipboard
   - Components: LoginForm, ApiKeyCreate, Dashboard

5. CI/CD 流水线重构
   - 重构 backend-ci.yml 为统一的 ci.yml
   - 前后端 4 个并行 Job + Postgres/Redis services
   - Race 检测、覆盖率收集与门禁、Docker 构建验证

6. E2E 自动化测试
   - e2e-test.sh 自动化脚本(Docker 启动→健康检查→测试→清理)
   - 用户注册→登录→API Key→网关调用完整链路测试
   - Mock 模式和 API Key 脱敏支持

7. 修复预存问题
   - tlsfingerprint dialer_test.go 缺失 build tag 导致集成测试编译冲突

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-02-08 12:05:39 +08:00