feat(llm): provider-abstract gateway with cost and failure controls

DeepSeek (OpenAI-compatible) as primary with a 智谱 GLM slot that is a real
routing path but raises until the 算法备案 model list is confirmed. Adds scene
routing, failover, a per-provider circuit breaker, structured output with
retry-on-validation-failure, prompts loaded by id+version from prompts/, and
cost attribution written to ob_ai_call for every attempt -- successes,
retries and failures alike.

Two controls exist because they are cheap now and expensive later: a
cache-prefix validator that rejects timestamps, uuids and nicknames in the
cached system prefix (a volatile prefix silently multiplies input cost), and
a hard platform-wide daily budget that rejects non-critical scenes instead of
merely alerting, since a retry storm outruns any human on-call.

packages/llm/providers/zhipu.py

@@ -0,0 +1,57 @@
+"""智谱 GLM provider — backup supplier slot (TD-04, 架构文档 §2 降级表).
+
+**Structure only.** The routing/failover code path is real: the gateway will
+route to this provider, the breaker will count its failures, and ``ob_ai_call``
+will record the attempt. Only the HTTP call is missing.
+
+Before implementing, resolve the open compliance item in 需求规格说明 §5.3:
+「备案覆盖的模型清单核实 —— 若备案未覆盖 DeepSeek + GLM 两家，双路由属超范围」.
+Shipping a working second route ahead of that check is a compliance risk, not a
+resilience win.
+"""
+
+from __future__ import annotations
+
+from packages.llm.errors import ProviderNotImplementedError
+from packages.llm.types import ProviderRequest, ProviderResponse
+
+_NOT_IMPLEMENTED_MESSAGE = (
+    "智谱 GLM provider is a structural placeholder and has no implementation. "
+    "To enable it: (1) confirm the 算法备案 model list covers GLM "
+    "(需求规格说明 §5.3), (2) implement complete() against "
+    "POST {base_url}/chat/completions, (3) map GLM usage fields onto "
+    "packages.llm.types.Usage, (4) set ZHIPU_API_KEY. "
+    "Until then the gateway treats this slot as a permanently failing provider."
+)
+
+
+class ZhipuGLMProvider:
+    def __init__(
+        self,
+        *,
+        api_key: str = "",
+        base_url: str = "https://open.bigmodel.cn/api/paas/v4",
+        default_model: str = "glm-4-flash",
+        name: str = "zhipu",
+    ) -> None:
+        self._api_key = api_key
+        self._base_url = base_url.rstrip("/")
+        self._default_model = default_model
+        self._name = name
+
+    @property
+    def name(self) -> str:
+        return self._name
+
+    @property
+    def default_model(self) -> str:
+        return self._default_model
+
+    async def complete(self, request: ProviderRequest) -> ProviderResponse:
+        raise ProviderNotImplementedError(self._name, _NOT_IMPLEMENTED_MESSAGE)
+
+    async def aclose(self) -> None:
+        return None
+
+
+__all__ = ["ZhipuGLMProvider"]