iomgaa/PolyGateway
1
0
Fork 0
Code Issues 3 Pull Requests Actions Packages Projects Releases 9 Wiki Activity

docs: sync env template, migration notes and architecture for M2.5

Browse Source
This commit is contained in:
iomgaa
2026-07-21 10:00:11 -04:00
parent 4c2a148db9
commit 910857fd13
3 changed files with 21 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
research-wiki/migrations/chsanalyzer.md
+10
View File
@@ -178,3 +178,13 @@ stack = ExtractionProviderStack(
| **G5** | ⚠️ 半开探针租约 TTL(探针持有者死亡后 TTL 过期自动可再探,scripts.py:96-105)与 `release_probe` 操作未见于 ARCH §7.4(只写单探针/epoch fencing);缺失则探针死锁 | M2 | **架构缺口**,修订 §7.4 |
| G6 | 装配期不变式守卫:`timeout_s*1000 ≤ lease TTL`(container.py:105-112)、`stall_window ≥ 最慢源 ttft/timeout`(container.py:115-124)须进库 from_env;契约中 settle/release **幂等性**须写进 §7.3 契约文字 | M2 | 缺口(轻),M2 设计补 |
| G7 | judge 默认 provider=anthropic:走实验室 OpenAI 兼容中转即可接入,若须直连 Anthropic 官方 API 则需新 transport(D2 有端口无里程碑)——迁移前与人类确认网关路径 | M3 | 决策项,非缺口 |
## M2.5 追记(2026-07-21,治理韧性设计的行为偏离)
| 偏离 | 对 CHS 迁移的影响 |
|---|---|
| 熔断增设失败率通道(窗口样本 ≥ min_calls 且失败率 ≥ fail_rate 即开路;429 不入两通道) | CHS 纯连续失败语义保留为通道 1;高失败率"半死源"将比 CHS 更早被隔离——P6 实证 CHS 语义对 10% 成功率源永不开路(病灶 1) |
| 开路时长指数递增,`retry_after_s` 上限由 cooldown_s(60s)变为 max_cooldown_s(缺省 300s) | **G1 消费点注意**: tracking.py arq defer 延期时长最多放大 5 倍;属期望行为(反复坏的源就该等更久) |
| 阈值自动抬升由全局并发改为源级并发 | CHS .env 注释约定的本意(防并发误熔)保留;全局并发抬升是 M2 错误移植,已废除 |
| 缺省选源 round_robin → health_aware(EWMA×在途 P2C) | 显式配置 `LLM__SELECTOR=round_robin` 者行为不变;迁移时建议直接吃新缺省 |