Commit Graph
15 Commits
Author SHA1 Message Date
wangbo 8362d6d27a feat(routing): 完善平台满载避让与故障轮转
为未配置并发上限的平台增加基于运行和等待任务数的软负载,并按非满载、有效优先级、缓存亲和与负载稳定排序。\n\n平台模型 RPM、TPM 和并发额度竞争失败时只轮转候选,不触发冷却、禁用或降级;用户组额度保持不可绕过。补齐异步冷却排队恢复、满载原因、选择原因与低基数指标。\n\n验证:go test ./...;go vet ./...;PostgreSQL 原子额度、准入队列、Worker 容量回收及故障策略 HTTP acceptance。
2026-08-03 17:42:34 +08:00
wangbo 3886048e0f fix(gemini): 移除空返图占位并暴露失败原因
Gemini 图片响应未包含可提取资源时,不再返回伪造占位图,改为非重试失败并保留安全拦截、候选结束状态和上游错误等结构化诊断。\n\n同步将诊断写入 HTTP 错误、任务结果、attempt 指标和日志,确保任务失败时不结算,并补充单元及 PostgreSQL 验收测试。\n\n验证:gofmt、go test ./... -count=1、go vet ./...、迁移安全检查、pnpm openapi。
2026-07-29 19:50:43 +08:00
wangbo 152885bbb6 fix(routing): 对齐缓存亲和力硬规则与审计 2026-07-24 23:51:27 +08:00
wangbo 8847d973a8 feat: add Responses API compatibility 2026-07-10 23:33:15 +08:00
wangbo 0e675e259c feat: add runner failover rules and cache affinity 2026-06-28 20:50:23 +08:00
wangbo 7f32446466 feat: support cached input token billing 2026-06-23 17:17:57 +08:00
wangbo 8ee7a7969e feat: expose retry priority in retry chain 2026-05-30 14:25:29 +08:00
wangbo 73c6d43e4b feat: filter candidates by request resolution 2026-05-18 15:50:08 +08:00
wangbo 90c3315468 fix(runner): record failed task attempts 2026-05-17 20:50:20 +08:00
wangbo 3225833f96 fix: align video generation payloads 2026-05-14 00:14:54 +08:00
wangbo c2696e7bbe feat(api): add load-aware client fallback 2026-05-12 16:59:51 +08:00
wangbo b9c9f457e9 feat: add parameter preprocessing audit trail 2026-05-12 13:54:51 +08:00
wangbo 05632172d0 feat: add runner failover policies and traces 2026-05-12 02:16:42 +08:00
wangbo 0049b246c1 feat: record task attempt chains 2026-05-11 22:39:45 +08:00
wangbo 53f8edfb67 feat: enrich task record details 2026-05-10 22:33:58 +08:00