GitLab

shaw authored Apr 07, 2026 and 陈曦 committed Apr 08, 2026

Anthropic近期引入基于system参数内容的第三方应用检测机制，原有的前置追加
Claude Code提示词策略无法通过检测（后续内容仍为非Claude Code格式触发429）。

新策略：对非Claude Code客户端的OAuth/SetupToken账号请求，将system字段
完整替换为Claude Code标识提示词，原始system内容作为user/assistant消息对
注入messages开头，模型仍接收完整指令。

仅影响/v1/messages路径，chat_completions和responses路径保持原有逻辑不变。
真正的Claude Code客户端请求完全不受影响（原样透传）。

shaw authored Apr 08, 2026 and 陈曦 committed Apr 08, 2026

commit f3aa54b7 的 rewriteSystemForNonClaudeCode 未能通过 Anthropic 第三方检测，
根因是两个关键信号与真实 Claude Code 不一致：

1. anthropic-beta 头缺少 claude-code-20250219：伪装路径主动将该 beta
   加入 drop set 并移除，但 Anthropic 依赖此 beta 识别 Claude Code 请求。
   修复：非 haiku 模型的伪装请求强制包含 claude-code beta。

2. system 字段使用 string 格式而非 array+cache_control：真实 Claude Code
   始终以 [{type,text,cache_control:{type:"ephemeral"}}] 发送 system，
   string 格式成为第三方检测信号。
   修复：rewriteSystemForNonClaudeCode 改为注入 array 格式。

附带调整：stripSystemCacheControl 按 system 是否被重写动态决定，
重写时保留 CC prompt 的 cache_control，未重写时（haiku/已含CC前缀）
保持原有剥离行为。

shaw authored Apr 07, 2026 and 陈曦 committed Apr 08, 2026

Anthropic近期引入基于system参数内容的第三方应用检测机制，原有的前置追加
Claude Code提示词策略无法通过检测（后续内容仍为非Claude Code格式触发429）。

新策略：对非Claude Code客户端的OAuth/SetupToken账号请求，将system字段
完整替换为Claude Code标识提示词，原始system内容作为user/assistant消息对
注入messages开头，模型仍接收完整指令。

仅影响/v1/messages路径，chat_completions和responses路径保持原有逻辑不变。
真正的Claude Code客户端请求完全不受影响（原样透传）。

Restore gateway_service.go, setting_handler.go, routes/admin.go,
dto/settings.go, group_repo.go, api_key_repo.go, wire_gen.go to
upstream/main versions and surgically remove only Sora references.

This preserves upstream-only features (RequireOauthOnly, RequirePrivacySet,
GroupResolution, etc.) that were missing when using release branch versions.

- applyRequestTierOverrides now uses filterValidIntervals consistently
  with applyTokenOverrides (per_request/image modes were not filtering)
- CostInput accepts optional pre-resolved pricing via Resolved field,
  eliminating duplicate Resolver.Resolve() calls in gateway billing paths

- Remove unused claudeMax*Tokens constants (Claude Max feature not included)
- Remove unused UsageMapHook type, SetUsageMapHook method, and usageToMap function
- Fix gofmt formatting in channel_service.go, openai_model_mapping_test.go,
  chatcompletions_to_responses.go

- Add int64(0) param to SelectAccountWithLoadAwareness callers (signature change from channel scheduling refactor)
- Add UsageMapHook type and struct field to StreamingProcessor
- Revert Claude Max cache billing code to upstream/main (not part of channel feature)
- Revert credits overages logic to upstream/main (non-channel change)
- Remove Instructions field reference (non-channel OpenAI feature)
- Restore sora_client_handler_test.go from upstream + add channel service nil params

- Change channel cache TTL from 60s to 10min (reduce unnecessary DB queries)
- Actively rebuild cache after CRUD instead of lazy invalidation
- Add slog.Warn logging for channel pricing restriction blocks (4 places)

P0-1: Credits degraded response retry + fail-open
- Add isAntigravityDegradedResponse() to detect transient API failures
- Retry up to 3 times with exponential backoff (500ms/1s/2s)
- Invalidate singleflight cache between retries
- Fail-open after exhausting retries instead of 5h circuit break

P1-1: Fix channel restriction pre-check timing conflict
- Swap checkClaudeCodeRestriction before checkChannelPricingRestriction
- Ensures channel restriction is checked against final fallback groupID

P1-2: Add interval pricing validation (frontend + backend)
- Backend: ValidateIntervals() with boundary, price, overlap checks
- Frontend: validateIntervals() with Chinese error messages
- Rules: MinTokens>=0, MaxTokens>MinTokens, prices>=0, no overlap

P2: Fix cross-platform same-model pricing/mapping override
- Store cache keys using original platform instead of group platform
- Lookup across matching platforms (antigravity→anthropic→gemini)
- Prevents anthropic/gemini same-name models from overwriting each other

- Fix 7 stale comments still mentioning "限制检查" in handlers/services
- Make billingModelForRestriction explicitly list channel_mapped case
- Add slog.Warn for error swallowing in ResolveChannelMapping and
  needsUpstreamChannelRestrictionCheck
- Document sticky session upstream check exemption

Move the model pricing restriction check from 8 handler entry points
to the account scheduling phase (SelectAccountForModelWithExclusions /
SelectAccountWithLoadAwareness), aligning restriction with billing:

- requested: check original request model against pricing list
- channel_mapped: check channel-mapped model against pricing list
- upstream: per-account check using account-mapped model

Handler layer now only resolves channel mapping (no restriction).
Scheduling layer performs pre-check for requested/channel_mapped,
and per-account filtering for upstream billing source.

- Extract resolveChannelPricing to DRY the resolver pattern shared by calculateImageCost/calculateTokenCost
- Remove unnecessary IIFE wrapper and pass accountRateMultiplier as parameter
- Extract resolveBillingMode, resolveMediaType, optionalSubscriptionID to simplify buildRecordUsageLog (104→65 lines)
- Extract shouldDeductAPIKeyQuota/shouldUpdateRateLimits/shouldUpdateAccountQuota methods on postUsageBillingParams to unify duplicated billing conditions

- Extract recordUsageCore with recordUsageOpts for parameterized differences
- RecordUsage (276 lines) → thin wrapper (~40 lines)
- RecordUsageWithLongContext (251 lines) → thin wrapper (~20 lines)
- Split billing logic into calculateSoraMediaCost, calculateImageCost,
  calculateTokenCost sub-functions
- Extract buildRecordUsageLog for usage log construction
- Net reduction: -79 lines, eliminated ~170 lines of duplication

- PricingSourceChannel/LiteLLM/Fallback for resolver source
- MediaTypeImage/Video/Prompt for result.MediaType
- Reuse BillingModeToken/BillingModeImage for billing mode
- Reuse BillingModelSourceChannelMapped/PlatformAnthropic in handler

Instead of hardcoding BillingMode="image" when ImageCount>0,
let cost.BillingMode (set by CalculateCostUnified/CalculateImageCost)
take priority. This ensures channel token pricing shows "token" mode.

When ImageCount > 0, check if channel has token pricing configured:
- YES (source=channel, mode=token) → use token billing with image_output_tokens
- NO → fall back to CalculateImageCost (original per-image billing)

This allows channels to configure $/MTok pricing for image generation
models while maintaining backward compatibility for setups without
channel pricing.

- Apply same nil-pointer protection as OpenAI path
- Remove unused accessToken/proxyURL params from checkAccountCredits

- Parse candidatesTokensDetails from Gemini API to separate image/text output tokens
- Add image_output_tokens and image_output_cost to usage_log (migration 089)
- Support per-image-token pricing via output_cost_per_image_token from model pricing data
- Channel pricing ImageOutputPrice override works in token billing mode
- Auto-fill image_output_price in channel pricing form from model defaults
- Add "channel_mapped" billing model source as new default (migration 088)
- Bills by model name after channel mapping, before account mapping
- Fix channel cache error TTL sign error (115s → 5s)
- Fix Update channel only invalidating new groups, not removed groups
- Fix frontend model_mapping clearing sending undefined instead of {}
- Credits balance precheck via shared AccountUsageService cache before injection
- Skip credits injection for accounts with insufficient balance
- Don't mark credits exhausted for "exhausted your capacity on this model" 429s

- 渠道模型映射：支持精确匹配和通配符映射，按平台隔离
- 渠道模型定价：支持 token/按次/图片三种计费模式，区间分层定价
- 模型限制：渠道可限制仅允许定价列表中的模型
- 计费模型来源：支持 requested/upstream 两种计费模型选择
- 用量统计：usage_logs 新增 channel_id/model_mapping_chain/billing_tier/billing_mode 字段
- Dashboard 支持 model_source 维度（requested/upstream/mapping）查看模型统计
- 全部 gateway handler 统一接入 ResolveChannelMappingAndRestrict
- 修复测试：同步 SoraGenerationRepository 接口、SQL INSERT 参数、scan 字段

- 4个缺失handler入口添加渠道映射+限制检查(ChatCompletions/Responses/Gemini)
- 模型限制错误信息优化，区分"模型不可用"和"无账号"
- OpenAI RecordUsage RequestedModel 改用 OriginalModel
- ResolveChannelMappingAndRestrict/ReplaceModelInBody 抽取到 ChannelService 消除跨service重复
- validateNoDuplicateModels 按 platform:model 去重
- 删除 Channel.ResolveMappedModel 死代码和 CalculateCostWithChannel Deprecated方法
- 移除冗余nil检查，抽取 validatePricingBillingMode 公共校验

- 抽取 ResolveChannelMappingAndRestrict 统一入口（5处→1个方法）
- 抽取 BuildModelMappingChain 到 ChannelMappingResult 方法（5处→1行调用）
- OpenAI 三入口 Forward 前应用渠道映射到请求体
- OpenAI Responses/Messages 限制检查添加错误响应
- 清理前端 3 处 console.log 调试日志

- GatewayService/OpenAIGatewayService 注入 ModelPricingResolver
- RecordUsage 从旧路径迁移到 CalculateCostUnified（支持 per_request/image 模式）
- 无渠道时自动回退旧路径，保持原有行为
- 长上下文双倍计费仅在无渠道定价时生效
- CostBreakdown 新增 BillingMode 字段，使用日志记录实际计费模式
- 模型限制错误改为与"无可用账号"相同的 503 响应

- DB: usage_logs 表新增 billing_mode VARCHAR(20) 列
- 后端: RecordUsage 写入时根据 image_count 判定计费模式
- 前端: 使用记录表格新增计费模式 badge 列 + 筛选下拉

- 缓存重构为 O(1) 哈希结构 (pricingByGroupModel, mappingByGroupModel)
- 渠道模型映射接入网关流程 (Forward 前应用, a→b→c 映射链)
- 新增 billing_model_source 配置 (请求模型/最终模型计费)
- usage_logs 新增 channel_id, model_mapping_chain, billing_tier 字段
- 每种计费模式统一支持默认价格 + 区间定价
- 渠道模型限制开关 (restrict_models)
- 分组按平台分类展示 + 彩色图标
- 必填字段红色星号 + 模型映射 UI
- 去除模型通配符支持

Cherry-picked from release/custom-0.1.106: a9117600

QTom authored Mar 31, 2026 and 陈曦 committed Apr 03, 2026

- buffer 公式从 baseRPM/5 改为 concurrency + maxSessions
  保留 baseRPM/5 作为 floor 向后兼容
- 粘性路径 fallback 新增 [StickyCacheMiss] 结构化日志
  reason: rpm_red / gate_check / session_limit / wait_queue_full / account_cleared
- session_limit 路径跳过 wait queue 重试（RegisterSession 拒绝无副作用）
- 典型配置 buffer 从 3 提升至 13，大幅减少高峰期 Prompt Cache Miss
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

QTom authored Mar 27, 2026 and 陈曦 committed Apr 03, 2026

为 OpenAI/Antigravity/Anthropic/Gemini 分组新增两个布尔控制字段：
- require_oauth_only: 创建/更新账号绑定分组时拒绝 apikey 类型加入
- require_privacy_set: 调度选号时跳过 privacy 未成功设置的账号并标记 error

后端：Ent schema 新增字段 + 迁移、Group CRUD 全链路透传、
      gateway_service 与 openai_account_scheduler 两套调度路径过滤
前端：创建/编辑表单 toggle 开关（OpenAI/Antigravity/Anthropic/Gemini 平台可见）
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

- buffer 公式从 baseRPM/5 改为 concurrency + maxSessions
  保留 baseRPM/5 作为 floor 向后兼容
- 粘性路径 fallback 新增 [StickyCacheMiss] 结构化日志
  reason: rpm_red / gate_check / session_limit / wait_queue_full / account_cleared
- session_limit 路径跳过 wait queue 重试（RegisterSession 拒绝无副作用）
- 典型配置 buffer 从 3 提升至 13，大幅减少高峰期 Prompt Cache Miss
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

为 OpenAI/Antigravity/Anthropic/Gemini 分组新增两个布尔控制字段：
- require_oauth_only: 创建/更新账号绑定分组时拒绝 apikey 类型加入
- require_privacy_set: 调度选号时跳过 privacy 未成功设置的账号并标记 error

后端：Ent schema 新增字段 + 迁移、Group CRUD 全链路透传、
      gateway_service 与 openai_account_scheduler 两套调度路径过滤
前端：创建/编辑表单 toggle 开关（OpenAI/Antigravity/Anthropic/Gemini 平台可见）
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

QTom authored Mar 27, 2026 and 陈曦 committed Mar 29, 2026

injectClaudeCodePrompt 和 systemIncludesClaudeCodePrompt 的 type switch
无法匹配 json.RawMessage 类型（Go typed nil 陷阱），导致 ForwardAsResponses
和 ForwardAsChatCompletions 路径中用户 system prompt 被替换为仅 Claude Code
banner。新增 normalizeSystemParam 将 json.RawMessage 转为标准 Go 类型。
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

shaw authored Mar 27, 2026 and 陈曦 committed Mar 29, 2026

新增功能：
- 新增 TLS 指纹 Profile CRUD 管理（Ent schema + 迁移 + Admin API + 前端管理界面）
- 支持账号绑定数据库中的自定义 TLS Profile，或随机选择（profile_id=-1）
- HTTPUpstream.DoWithTLS 接口从 bool 改为 *tlsfingerprint.Profile，支持按账号指定 Profile
- AccountUsageService 注入 TLSFingerprintProfileService，统一 usage 场景与网关的 Profile 解析逻辑

代码优化：
- 删除已被 TLSFingerprintProfileService 完全取代的 registry.go 死代码（418 行）
- 提取 3 个 dialer 的重复 TLS 握手逻辑为 performTLSHandshake() 共用函数
- 修复 GetTLSFingerprintProfileID 缺少 json.Number 处理的 bug
- gateway_service.Forward 中 ResolveTLSProfile 从重试循环内重复调用改为预解析局部变量
- 删除冗余的 buildClientHelloSpec() 单行 wrapper 和 int64(e.ID) 无效转换
- tls_fingerprint_profile_cache.go 日志从 log.Printf 改为 slog 结构化日志
- dialer_capture_test.go 添加 //go:build integration 标签，防止 CI 失败
- 去重 TestProfileExpectation 类型至共享 test_types_test.go
- 修复 9 个测试文件缺少 tlsfingerprint import 的编译错误
- 修复 error_policy_integration_test.go 中 handleError 回调签名被错误替换的问题

injectClaudeCodePrompt 和 systemIncludesClaudeCodePrompt 的 type switch
无法匹配 json.RawMessage 类型（Go typed nil 陷阱），导致 ForwardAsResponses
和 ForwardAsChatCompletions 路径中用户 system prompt 被替换为仅 Claude Code
banner。新增 normalizeSystemParam 将 json.RawMessage 转为标准 Go 类型。
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

新增功能：
- 新增 TLS 指纹 Profile CRUD 管理（Ent schema + 迁移 + Admin API + 前端管理界面）
- 支持账号绑定数据库中的自定义 TLS Profile，或随机选择（profile_id=-1）
- HTTPUpstream.DoWithTLS 接口从 bool 改为 *tlsfingerprint.Profile，支持按账号指定 Profile
- AccountUsageService 注入 TLSFingerprintProfileService，统一 usage 场景与网关的 Profile 解析逻辑

代码优化：
- 删除已被 TLSFingerprintProfileService 完全取代的 registry.go 死代码（418 行）
- 提取 3 个 dialer 的重复 TLS 握手逻辑为 performTLSHandshake() 共用函数
- 修复 GetTLSFingerprintProfileID 缺少 json.Number 处理的 bug
- gateway_service.Forward 中 ResolveTLSProfile 从重试循环内重复调用改为预解析局部变量
- 删除冗余的 buildClientHelloSpec() 单行 wrapper 和 int64(e.ID) 无效转换
- tls_fingerprint_profile_cache.go 日志从 log.Printf 改为 slog 结构化日志
- dialer_capture_test.go 添加 //go:build integration 标签，防止 CI 失败
- 去重 TestProfileExpectation 类型至共享 test_types_test.go
- 修复 9 个测试文件缺少 tlsfingerprint import 的编译错误
- 修复 error_policy_integration_test.go 中 handleError 回调签名被错误替换的问题

新增独立开关控制 API Key 账号的签名整流功能，支持配置自定义
匹配关键词以捕获不同格式的上游错误响应。

- 新增 apikey_signature_enabled 开关（默认关闭）
- 新增 apikey_signature_patterns 自定义关键词配置
- 内置签名检测规则对 API Key 账号同样生效
- 自定义关键词对完整响应体做不区分大小写匹配
- 重试二阶段检测仅做模式匹配，不重复校验开关
- Handler 层校验关键词数量（≤50）和长度（≤500）
- API 响应 nil patterns 统一序列化为空数组
- OAuth/SetupToken/Upstream/Bedrock 账号行为不变