pi-ccswitch-auto-switch 0.3.1 → 0.3.2

package/README.md

@@ -13,7 +13,7 @@ The extension observes real Pi requests, records sanitized health signals, and
 - Endpoint circuit breakers for DNS, connection, server, and streaming failures.
 - Endpoint platform isolation: when several models on the same endpoint (same BaseURL/provider) fail within one round, the whole endpoint is isolated so sibling models from the same platform are not tried one by one.
 - Model-only isolation for missing models and incompatible parameters.
-- Learns model-family content-policy constraints from real failures (`glm-4.5` and `GLM-4.6`, for example, are both `glm`) and avoids the entire family during current and future content-policy failover chains. Evidence expires after 30 days.
+- Learns model-family content-policy constraints from real failures (`glm-4.5` and `GLM-4.6`, for example, are both `glm`) and avoids the entire family during the current content-policy failover chain. Evidence is session-scoped: it is cleared at `session_start`, because a new session may handle tasks that are not content-sensitive, so the family becomes eligible again until this session observes its own content-policy rejection.
 - Applies hard health, failure-domain, policy-family, input-modality, and context filters before ranking candidates by provider diversity, model equivalence, capability compatibility, and historical reliability.
   - Independent-provider semantics: a copied vendor (e.g. `my-provider-copy`) shares the BaseURL but is a distinct provider with its own endpoint key, so isolating one never blocks the other.
 - 90-second first-response and 120-second streaming-idle watchdogs.
@@ -107,7 +107,7 @@ The state machine waits for Pi's native retry cycle to settle before switching.
 | `429` | Provider | `Retry-After` when present; otherwise 5 minutes |
 | DNS, network, `408`, `5xx`, interrupted stream | Endpoint | 2 minutes |
 | `404`, invalid model, incompatible parameters | Model | 15 minutes |
-| Content-policy or sensitive-content rejection | Current model + learned model-family constraint | 2-minute model cooldown; 30-day family evidence |
+| Content-policy or sensitive-content rejection | Current model + learned model-family constraint | 2-minute model cooldown; family evidence is session-scoped (cleared at `session_start`) |
 | Context overflow | Current round only | None |
 
 Cooldowns grow exponentially within bounded limits. Learned policy-constrained families are excluded only after a content-policy rejection; they remain eligible during ordinary rate-limit, network, and model-configuration failovers. Context-overflow retries only consider models with a larger context window, and image requests do not move to a model that explicitly supports text only. Unknown errors are isolated to the current model and still fail over normally. A user cancellation is the only aborted turn that does not trigger failover; watchdog cancellations are recorded as timeouts.

package/README.zh-CN.md

@@ -12,7 +12,7 @@
 - 对认证、额度、账单和限流实施 Provider 优先熔断。
 - 对 DNS、连接、服务端和流中断实施端点熔断。
 - 模型不存在或参数不兼容时仅隔离该模型。
-- 内容审查失败会学习模型系列约束（例如 `glm-4.5`、`GLM-4.6` 同属 `glm`），本轮及后续内容审查故障转移会避开整个受限系列；证据 30 天后自动过期。
+- 内容审查失败会学习模型系列约束（例如 `glm-4.5`、`GLM-4.6` 同属 `glm`），在本轮内容审查故障转移中避开整个受限系列。该约束按 session 隔离：`session_start` 时清除，因为新 session 处理的任务不一定涉及审查内容，模型系列恢复可用，直到本 session 再次观察到内容审查拒绝。
 - 候选模型先经过健康、故障域、审查系列、输入模态和上下文窗口硬过滤，再按跨 Provider、模型等价性、能力兼容和历史稳定性排序。
 - 首响应 90 秒、流式停滞 120 秒看门狗。
 - 指数退避、`Retry-After` 支持和跨进程 half-open 单探针租约。
@@ -95,7 +95,7 @@ CCS v0.3.1 ✓70/131 · ⏳61 · ⛔0 · 🔄3 · provider/model-id
 | `429` | Provider | 优先使用 `Retry-After`，否则 5 分钟 |
 | DNS、网络、`408`、`5xx`、流中断 | 端点 | 2 分钟 |
 | `404`、模型不存在、参数不兼容 | 模型 | 15 分钟 |
-| 内容审查/敏感拦截 | 当前模型 + 模型系列约束 | 模型冷却 2 分钟；系列避让证据 30 天 |
+| 内容审查/敏感拦截 | 当前模型 + 模型系列约束 | 模型冷却 2 分钟；系列约束按 session 隔离（`session_start` 时清除） |
 | 上下文溢出 | 仅本轮 | 无 |
 
 冷却时间会指数增长但有上限。发生内容审查时，插件只在审查故障转移链中避开已标记系列，普通限流、网络或模型配置故障仍可选择这些模型。上下文溢出时只会选择上下文窗口更大的模型；带图片的请求不会切到明确仅支持文本的模型。无法归类的异常按单模型故障处理并正常切换。用户主动取消是唯一不会触发故障转移的异常终止；看门狗取消会记录为超时。

package/health.ts

@@ -158,8 +158,9 @@ export class HealthStore {
   }
 
   /**
-   * 内容审查约束比限流/网络故障稳定得多，但仍设置 30 天证据有效期，避免模型策略升级后
-   * 永久污染路由。再次观察到同系列审查会增加证据计数并刷新有效期。
+   * 内容审查约束是当前 session 的观测，按 session 隔离：新 session 可能处理
+   * 不涉及审查内容的任务，因此约束在 session_start 时清除，不跨 session 继承；
+   * 本 session 内再次观察到同系列审查会增加证据计数并刷新时间戳。
    */
   recordContentPolicyConstraint(family: string, model: ModelRef, message?: string): void {
     const constraints = this.state.contentPolicyFamilies ??= {}
@@ -235,6 +236,15 @@ export class HealthStore {
     this.touch()
   }
 
+  /**
+   * 清除已学习的模型系列内容审查约束。新 session 可能处理不同任务，不假设
+   * 对审查内容敏感，因此不在 session 间继承审查约束。
+   */
+  clearContentPolicyConstraints(): void {
+    this.state.contentPolicyFamilies = {}
+    this.touch()
+  }
+
   isBlocked(model: ModelRef, now = Date.now()): boolean {
     return [this.state.models[modelKey(model)], this.state.providers[model.provider], this.state.endpoints[endpointKey(model)]]
       .some(record => Boolean(record?.disabled || (record?.cooldownUntil && record.cooldownUntil > now) || (record?.leaseUntil && record.leaseUntil > now)))

package/index.ts

@@ -8,7 +8,7 @@ const STREAM_IDLE_TIMEOUT = 120_000
 const MAX_ATTEMPTS = 5
 const ROUND_LIMIT = 8 * 60_000
 const RPC_PROTOCOL_VERSION = 1
-const EXTENSION_VERSION = '0.3.1'
+const EXTENSION_VERSION = '0.3.2'
 // 同端点（BaseURL 相同）连续失败达到该次数即隔离该端点，避免同一个平台的多个模型逐个试错耗尽本轮切换
 const ENDPOINT_FAIL_THRESHOLD = 3
 
@@ -267,7 +267,16 @@ export default function (pi: ExtensionAPI) {
     await exhaust(ctx, '没有健康的候选模型')
   }
 
-  pi.on('session_start', async (_event, ctx) => { await health.load(); sessionSwitches = 0; await refresh(ctx); await health.log('extension started') })
+  pi.on('session_start', async (_event, ctx) => {
+    await health.load()
+    sessionSwitches = 0
+    // 新 session 可能处理不涉及审查内容的任务，上一 session 学到的模型系列审查约束
+    // 不跨 session 继承；若本 session 再次遇到 content_policy 失败会重新学习并在本轮避开。
+    health.clearContentPolicyConstraints()
+    await health.flush()
+    await refresh(ctx)
+    await health.log('extension started')
+  })
   pi.on('session_shutdown', async (_event, ctx) => { clearWatchdog(); ctx.ui.setWorkingMessage(); if (sessionSwitches > 0) await health.log(`session ended with ${sessionSwitches} successful switches`); await health.flush() })
   pi.on('input', (event, ctx) => {
     if (event.source === 'extension') return { action: 'continue' }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-ccswitch-auto-switch",
-  "version": "0.3.1",
+  "version": "0.3.2",
   "description": "Provider-first automatic model failover extension for Pi and CC Switch",
   "license": "MIT",
   "keywords": [