GitLab

对齐/v1/responses的Forward方法，修复两处不一致：
- 成功响应时从响应头提取OAuth账号的Codex使用量数据
- 非failover错误场景下应用管理员配置的错误透传规则

fix(usage): 修复用量查询 429 重试风暴，增加负缓存、请求去重与随机延迟

- Fix gofmt: align struct field comments in UsageCache, trim trailing
  whitespace on const comments
- Fix errcheck: use comma-ok on type assertion for singleflight result
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Prevents 429 rate-limit retry storms and reduces upstream correlation risk
for Anthropic usage API queries.

Three changes:
1. Negative caching (1 min TTL) — 429/error responses are now cached,
   preventing every subsequent page load from re-triggering failed API calls.
2. singleflight dedup — concurrent requests for the same account are
   collapsed into a single upstream call, preventing cache stampede.
3. Random jitter (0–800 ms) — staggers multi-account cache-miss bursts so
   requests from different accounts don't hit upstream simultaneously with
   identical TLS fingerprints, reducing anti-abuse correlation risk.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

fix: Setup Token 账号使用真实 utilization 值替代状态估算

fix(openai): 修复 WS 模式下用量窗口不显示

从响应头 anthropic-ratelimit-unified-5h-utilization 获取并存储真实
utilization 值，解决进度条始终显示 0% 的问题。窗口重置时清除旧值，
避免残留上个窗口的数据。
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

feat(openai): 添加 /v1/messages 端点和 API 兼容层

fix(openai): support remote compact task

fix(openai): 统一专属倍率计费链路并补齐回归测试

fix: 修复账号列表五小时用量显示为 $0.00 的问题

移除订阅扣费 stub 中未被使用的状态字段与赋值，
消除 golangci-lint 的 unused 告警，保持回归测试语义不变。
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

抽取共享的用户分组专属倍率解析器，统一缓存、singleflight 与回退逻辑。\n\n让 OpenAI 独立计费链路复用专属倍率解析，修复 usage 记录与实际扣费未命中用户专属倍率的问题。\n\n补齐 OpenAI 计费与解析器单元测试，并修复全量回归中暴露的 lint 阻塞项。\n\nCo-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

commit 80ae592c 引入 lite 模式优化首次加载性能，但将窗口费用查询也一起跳过了。
commit 491a7444

 尝试用 30 秒快照缓存修复，但缓存过期后问题复现。

移除窗口费用查询的 lite/非 lite 区分，始终执行 PostgreSQL 聚合查询。
同时删除不再需要的 account_window_cost_cache.go 文件。
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

在 admin.proxies 下添加 protocols 对象的中文翻译，
与 en.ts 保持一致。
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

- Add recoverAnthropicMessagesPanic for Messages handler to return
  Anthropic-formatted errors instead of OpenAI Responses format on panic
- Add nil check for rateLimitService.HandleUpstreamError in
  ForwardAsAnthropic to match defensive pattern used elsewhere
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Add Anthropic Messages API support for OpenAI platform groups, enabling
clients using Claude-style /v1/messages format to access OpenAI accounts
through automatic protocol conversion.

- Add apicompat package with type definitions and bidirectional converters
  (Anthropic ↔ Chat, Chat ↔ Responses, Anthropic ↔

 Responses)
- Implement /v1/messages endpoint for OpenAI gateway with streaming support
- Add model mapping UI for OpenAI OAuth accounts (whitelist + mapping modes)
- Support prompt caching fields and codex OAuth transforms
- Fix tool call ID conversion for Responses API (fc_ prefix)
- Ensure function_call_output has non-empty output field
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

fix(openai): passthrough accounts bypass model mapping check

透传模式账号仅替换认证，应允许所有模型通过。之前调度阶段的
isModelSupportedByAccount 不感知透传模式，导致 model_mapping
中未配置的新模型（如 gpt-5.4）被拒绝返回 503。

feat(openai): 增加 GPT-5.4 支持并修复长上下文计费与白名单回归

lite 模式下从快照缓存读取窗口费用，非 lite 模式查询后写入缓存

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)

- 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)

feat: add account load_factor for scheduling load calculation

feat: add configurable spending limit for API Key accounts

LoadFactor should only affect scheduling weight, not load rate reporting.

Replace `<` comparisons with Math.max/ternary+>= to avoid Vue template
parser treating `<` as HTML tag start in attribute values.

- Fix bulk edit: send 0 instead of null/NaN to clear load_factor
- Fix edit modal: explicit NaN check instead of implicit falsy
- Fix create modal: use ?? instead of || for load_factor
- Add load_factor upper limit validation (max 10000)
- Add //go:build unit tag and self-contained intPtrHelper in test
- Add design intent comments on WaitPlan.MaxConcurrency