On PostgreSQL the logs table is a RANGE-partitioned table on created_at
owned by pg_partman; GORM AutoMigrate would alter its composite primary
key (id, created_at) and create redundant per-partition indexes,
breaking startup. Extract ensureLogTable() to skip AutoMigrate on
PostgreSQL (existence check only) while keeping the original behavior
on SQLite/MySQL. The dedicated LOG_SQL_DSN PostgreSQL branch follows
the same rule. Ship the matching postgres-partman image (Dockerfile +
01-partition.sql) that creates and maintains the partitioned table via
pg_partman + pg_cron.

Co-Authored-By: ZCode <noreply@anthropic.com>

Co-Authored-By: Codex <noreply@anthropic.com>

Co-Authored-By: Codex <noreply@anthropic.com>

Co-Authored-By: Codex <noreply@anthropic.com>

Co-Authored-By: Codex <noreply@anthropic.com>

Co-Authored-By: Codex <noreply@anthropic.com>

Add two layers of defense:

1. Webhook availability guard (controller/payment_webhook_availability.go):
   - isStripeWebhookEnabled() checks StripeWebhookSecret != "" before processing
   - isCreemWebhookEnabled(), isWechatPayWebhookEnabled(), isAlipayWebhookEnabled()
   - isEpayWebhookEnabled() with similar checks for all payment webhooks
   - Applied to: StripeWebhook, CreemWebhook, WechatPayWebhook, AlipayPayWebhook,
     EpayNotify, SubscriptionEpayNotify

2. PaymentProvider field (model/topup.go):
   - New PaymentProvider field on TopUp to identify which gateway created the order
   - Recharge() checks PaymentProvider == PaymentProviderStripe
   - rechargeByQRCodePayment() checks PaymentProvider matches wechat/alipay
   - RechargeCreem() checks PaymentProvider == PaymentProviderCreem
   - All payment controllers set PaymentProvider when creating orders

Root cause: When StripeWebhookSecret was empty, ComputeSignature used
an empty HMAC key, allowing attackers to forge valid signatures and
complete orders from any payment gateway without actually paying.

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

Co-Authored-By: Codex <noreply@anthropic.com>

Co-Authored-By: Codex <noreply@anthropic.com>

Co-Authored-By: Codex <noreply@anthropic.com>

Add KlingAiping usage-token settlement from upstream final_unit_deduction and duration fallback handling.

Normalize model endpoint config persistence and convert model pricing values between stored USD and display currency in the editor.

Co-Authored-By: Codex <noreply@anthropic.com>

Co-Authored-By: Codex <noreply@anthropic.com>

- 避免仅编辑展示价格时写入默认真实计费规则
- 将 video-extend 纳入 KlingAiping 原生渠道约束
- 补充相关保存决策、渠道约束和计费展示测试

Co-Authored-By: Codex <noreply@anthropic.com>

- Asset: /api/v1/volcengine/asset → /api/v1/multimodal/sd/assets
- Create video task: /api/v1/videos → /api/v1/multimodal/sd/videos/contents/generations/tasks
- Query video task: /api/v1/videos/{id} → /api/v1/multimodal/sd/videos/contents/generations/tasks/{id}

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

- 新增 KlingAiping 渠道 (ChannelTypeKlingAiping = 59)
- 新增 KlingAiping 原生路由 (text2video/image2video/omni-video 等)
- 新增 KlingAiping adaptor (relay/channel/task/kling/aiping/)
- 前端模型定价页面新增「真实计费 / 展示价格」Tab 切换
- 新增 DisplayPricing 组件，支持模型展示价格的增删改
- 新增 displayPricingConfig 工具函数及测试
- 修复 KlingAiping 路由路径匹配与 task 错误响应

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

新建用户时异步同步到海外 slave 节点，确保跨区域用户数据一致。
补充前两个 commit 全部新功能的中英文翻译字符串。

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

前端新增 UsageLogExpandedDetail 点击展开面板和 BillingDetailSummary
hover 计费详情，后端修复 per_1m_tokens 任务结算日志的 completion_tokens
从零值改为写入 total_tokens 实际值。新增 taskBillingLogMerge 合并
任务预扣/结算日志。

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

前端模型定价配置 UI 增强（Editor/GeneratorModal/PreviewPanel/PricingGrid），
后端新增 representativeMatrixPrice 取矩阵定价表中位数作为展示价格，
替代原来的 model_price=0 导致的价格展示为空问题。

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

- feat: video multidimensional matrix pricing system
- feat: doubao aiping video channel support
- feat: relay capture JSON storage with header redaction
- feat: SSE stream error handling (Gemini/OpenAI/Responses)
- test: E2E test infrastructure (Playwright + Chromium)

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

- relay: SSE stream error exposure improvements (Claude/Gemini/OpenAI)
- test: add mock capture server for E2E relay capture testing

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

- Gemini stream: 捕获 promptFeedback.blockReason 后停止流并返回 OpenAPI 错误
- OpenAI stream: 检测 SSE 帧中的 error 响应，在首帧写入前转换为 API 错误
- Responses stream: 处理 response.error / response.failed / error 类型帧
- 新增 ClearEventStreamHeadersIfNotWritten 工具方法，错误转换时清除流式头
- 新增 streamOpenAIErrorFromData，从 SSE data 帧提取 OpenAPI 错误结构
- 上游错误体会截断后存入 UpstreamBody 供排障

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

Co-Authored-By: Codex <noreply@anthropic.com>

Co-Authored-By: Codex <noreply@anthropic.com>

Co-Authored-By: Codex <noreply@anthropic.com>

Co-Authored-By: Codex <noreply@anthropic.com>

Co-Authored-By: Codex <noreply@anthropic.com>

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