fix(llm): preserve serving retry policy

docs/core-data-structures/llm-streaming.i18n.yaml

@@ -2,5 +2,5 @@
 # side as of the last confirmed-consistent state. Both languages carry equal authority;
 # after editing either side, bring the other along and re-record with:
 #   pnpm run verify-translation-pairing --write
-llm-streaming.md: 2917d9796b32956f11ed5703f1f34c7510d0526c
-llm-streaming.zh.md: 1bf7b3c226b5878deff3252396d96c2e97bf2982
+llm-streaming.md: 2d185e694f272114d4efa4f6be9020ef7c9a950f
+llm-streaming.zh.md: 9734c643724e19fc39aff2ee94bd8de61935e16c

docs/core-data-structures/llm-streaming.md

@@ -59,7 +59,7 @@ Every adapter MUST obey these, and every consumer may rely on them:
 
 - **`usage` before `finish`, nothing after `finish`.** Defer both to the provider's end-of-stream marker so a trailing usage-only chunk can't violate the ordering.
 - **Tool-call `arguments` stay raw JSON strings end-to-end.** Partial fragments stream via `argumentsDelta`; a provider that hands back parsed objects re-stringifies at `block-end`.
-- **Two sanctioned error paths, one fact shape.** A failure may either THROW from `stream()` (transport/protocol errors) **or** end the stream with `finish {kind:'error'|'aborted', failure}` (provider in-band errors, for adapters that can't throw mid-stream). `LlmError.failure` carries the same `LlmFailure`. The final adapter boundary preserves the exact thrown `Error` object and associates immutable facts with that call; the agent loop closes the failed step and offers the error, facts, and immutable prior-retried facts to `agent/request-error`. Absent recovery the structured failure becomes the turn error, and no normal assistant message or tool side effect is committed for that attempt.
+- **Two sanctioned error paths, one fact shape.** A failure may either THROW from `stream()` (transport/protocol errors) **or** end the stream with `finish {kind:'error'|'aborted', failure}` (provider in-band errors, for adapters that can't throw mid-stream). `LlmError.failure` carries the same `LlmFailure`. The final adapter boundary preserves the exact thrown `Error` object and associates immutable facts plus the serving registration's immutable retry policy with that call; the agent loop closes the failed step and offers the error, facts, immutable prior-retried facts, and serving policy to `agent/request-error`. Absent recovery the structured failure becomes the turn error, and no normal assistant message or tool side effect is committed for that attempt.
 - **One adapter call is one provider attempt.** Adapters disable library retries. Agent-level recovery opens another durable numbered step; direct `ctx.llm.stream()` callers remain single-attempt.
 - **Provider stalls are bounded at the transport.** Both shipping remote adapters expose positive finite `streamIdleTimeoutMs` with a five-minute default. The watchdog arms only while iterator `next()` is outstanding, uses one stable signal for the whole request, maps its own expiry to `TIMEOUT`, and keeps an earlier caller abort as `ABORTED`.
 - **Context overflow has one canonical code.** Both DeepSeek adapters classify explicit provider detail through `isContextWindowExceededError()` and surface `CONTEXT_WINDOW_EXCEEDED`, whether the failure arrives as a thrown HTTP `LlmError` or an in-band finish error. Consumers route on the code, never provider text.
@@ -70,7 +70,7 @@ This contract is pinned down by two deliberately independent implementations: `d
 
 ## `ResolvedRetryPolicy`
 
-Provider configuration resolves before route registration into an immutable discriminated union. Normal mode carries `mode: 'normal'`, finite `maxRetries`, `retryableCodes`, and required `initialDelayMs`, `maxDelayMs`, and `jitterRatio`; always mode carries `mode: 'always'` and the same required backoff fields without a finite maximum. `LlmService.providerRetryPolicy(provider)` returns the captured value and supplies normal defaults when the adapter omits one. The [generated config catalog](../config-catalog.md) owns the optional input shapes.
+Provider configuration resolves before route registration into an immutable discriminated union. Normal mode carries `mode: 'normal'`, finite `maxRetries`, `retryableCodes`, and required `initialDelayMs`, `maxDelayMs`, and `jitterRatio`; always mode carries `mode: 'always'` and the same required backoff fields without a finite maximum. `LlmService.providerRetryPolicy(provider)` returns the currently registered value and supplies normal defaults when the adapter omits one; `llmRetryPolicyOf(stream)` returns the exact serving registration's captured value after that call enters its final adapter boundary, so later route disposal or replacement cannot change an in-flight failure's recovery policy. The [generated config catalog](../config-catalog.md) owns the optional input shapes.
 
 ## `AppIdentity` — app attribution
 

docs/core-data-structures/llm-streaming.zh.md

@@ -59,7 +59,7 @@ interface LlmFailure {
 
 - **`usage` 在 `finish` 之前，`finish` 之后不再有任何分片。** 将两者都推迟到提供方的流结束标记，这样尾部的 usage-only 分片就不会违反顺序。
 - **工具调用的 `arguments` 全程保持原始 JSON 字符串。** 部分片段通过 `argumentsDelta` 流式传输；如果提供方返回的是已解析的对象，适配器在 `block-end` 时重新序列化为字符串。
-- **两条受支持的错误路径，一种事实形状。** 失败可以从 `stream()` 抛出（传输/协议错误），**或者**以 `finish {kind:'error'|'aborted', failure}` 结束流（无法在流中途抛异常的适配器用它表示提供方带内错误）。`LlmError.failure` 携带同一个 `LlmFailure`。最终适配器边界保留被抛出的确切 `Error` 对象，并将不可变事实关联到该调用；agent loop（智能体循环）关闭失败的步骤，再把错误、事实与不可变的先前已重试事实提供给 `agent/request-error`。若未恢复，结构化失败会成为轮次错误，并且该次尝试不会提交正常 assistant 消息或工具副作用。
+- **两条受支持的错误路径，一种事实形状。** 失败可以从 `stream()` 抛出（传输/协议错误），**或者**以 `finish {kind:'error'|'aborted', failure}` 结束流（无法在流中途抛异常的适配器用它表示提供方带内错误）。`LlmError.failure` 携带同一个 `LlmFailure`。最终适配器边界保留被抛出的确切 `Error` 对象，并将不可变事实以及实际提供服务的注册项所持不可变重试策略关联到该调用；agent loop（智能体循环）关闭失败的步骤，再把错误、事实、不可变的先前已重试事实与实际提供服务的策略提供给 `agent/request-error`。若未恢复，结构化失败会成为轮次错误，并且该次尝试不会提交正常 assistant 消息或工具副作用。
 - **一次适配器调用就是一次提供方尝试。** 适配器禁用库重试。agent 层恢复会打开另一个持久、带编号的步骤；直接调用 `ctx.llm.stream()` 的调用方仍然只尝试一次。
 - **提供方停顿在传输层受到时限约束。** 两个已交付的远程适配器都暴露正数且有限的 `streamIdleTimeoutMs`，默认五分钟。watchdog 只在 iterator `next()` 尚未完成时启动，整个请求使用同一个稳定 signal，把自身到期映射为 `TIMEOUT`，并把更早发生的调用方中止保留为 `ABORTED`。
 - **上下文溢出只有一个规范 code。** 两个 DeepSeek 适配器都通过 `isContextWindowExceededError()` 对提供方的显式细节分类并暴露 `CONTEXT_WINDOW_EXCEEDED`，无论失败以抛出的 HTTP `LlmError` 还是带内 finish error 到达。消费方按 code 路由，绝不依赖提供方文本。
@@ -70,7 +70,7 @@ interface LlmFailure {
 
 ## `ResolvedRetryPolicy`
 
-提供方配置会在路由注册前解析为不可变的可辨识联合类型。normal 模式包含 `mode: 'normal'`、有限的 `maxRetries`、`retryableCodes`，以及必填的 `initialDelayMs`、`maxDelayMs` 和 `jitterRatio`；always 模式包含 `mode: 'always'` 和相同的必填退避字段，但不含有限上限。`LlmService.providerRetryPolicy(provider)` 返回捕获的值；适配器未提供策略时，该方法会补上 normal 默认值。可选输入形状由[生成的配置目录](../config-catalog.md)定义。
+提供方配置会在路由注册前解析为不可变的可辨识联合类型。normal 模式包含 `mode: 'normal'`、有限的 `maxRetries`、`retryableCodes`，以及必填的 `initialDelayMs`、`maxDelayMs` 和 `jitterRatio`；always 模式包含 `mode: 'always'` 和相同的必填退避字段，但不含有限上限。`LlmService.providerRetryPolicy(provider)` 返回当前已注册的值；适配器未提供策略时，该方法会补上 normal 默认值。调用进入最终适配器边界后，`llmRetryPolicyOf(stream)` 返回实际提供服务的确切注册项所捕获的值，因此后续路由 dispose 或替换无法改变请求进行期间发生的失败所用恢复策略。可选输入形状由[生成的配置目录](../config-catalog.md)定义。
 
 ## `AppIdentity`：应用归属