feat(awooop): surface ai provider route status

apps/api/src/api/v1/platform/operator_runs.py

@@ -25,6 +25,9 @@ from src.core.awooop_operator_auth import (
 from src.services.platform_operator_service import (
     decide_approval as decide_approval_svc,
 )
+from src.services.platform_operator_service import (
+    get_ai_route_status as get_ai_route_status_svc,
+)
 from src.services.platform_operator_service import (
     get_awooop_status_chain as get_awooop_status_chain_svc,
 )
@@ -99,6 +102,21 @@ class ListCallbackRepliesResponse(BaseModel):
     per_page: int
 
 
+class AiRouteStatusResponse(BaseModel):
+    schema_version: str
+    workload_type: str
+    policy_order: list[dict[str, Any]]
+    selected_provider: str | None = None
+    selected_url: str | None = None
+    selected_model: str | None = None
+    fallback_chain: list[dict[str, Any]]
+    route_reason: str
+    route_source: str
+    route_error: str | None = None
+    health: dict[str, dict[str, Any]]
+    checked_at: datetime
+
+
 class ApprovalItem(BaseModel):
     run_id: UUID
     project_id: str
@@ -198,6 +216,24 @@ async def list_callback_replies(
     )
 
 
+@router.get(
+    "/ai-route-status",
+    response_model=AiRouteStatusResponse,
+    summary="查詢 AI Provider 路由狀態",
+    description=(
+        "回傳目前 Ollama/Gemini 路由策略、即時 primary、fallback chain 與健康狀態；"
+        "只讀，不觸發推理或自動修復。"
+    ),
+)
+async def get_ai_route_status(
+    workload_type: str | None = Query(
+        "deep_rca",
+        description="工作負載類型，例如 deep_rca/hermes/interactive/embedding/rag/code_review/image_analysis",
+    ),
+) -> dict[str, Any]:
+    return await get_ai_route_status_svc(workload_type=workload_type)
+
+
 @router.get(
     "/runs/{run_id}/detail",
     summary="查詢 Run 詳細時間線",

apps/api/src/services/platform_operator_service.py

@@ -13,7 +13,7 @@ import uuid
 from collections import defaultdict
 from collections.abc import Mapping
 from datetime import UTC, datetime
-from typing import Any
+from typing import Any, get_args
 from uuid import UUID
 
 import structlog
@@ -38,6 +38,17 @@ from src.services.awooop_truth_chain_service import (
     _summarize_mcp,
     fetch_truth_chain,
 )
+from src.services.ollama_endpoint_resolver import (
+    OllamaEndpointSelection,
+    OllamaWorkloadType,
+    resolve_ollama_order,
+)
+from src.services.ollama_failover_manager import (
+    OllamaEndpoint,
+    OllamaRoutingResult,
+    get_ollama_failover_manager,
+)
+from src.services.ollama_health_monitor import HealthReport
 from src.services.run_state_machine import transition
 
 logger = structlog.get_logger(__name__)
@@ -74,6 +85,8 @@ _CALLBACK_REPLY_RAW_STATUS_BY_FILTER = {
     "failed": "callback_reply_failed",
 }
 _CALLBACK_REPLY_ACTION_RE = re.compile(r"^[a-z0-9_:-]{1,64}$", re.IGNORECASE)
+_AI_ROUTE_STATUS_SCHEMA_VERSION = "awooop_ai_route_status_v1"
+_AI_ROUTE_WORKLOADS = set(get_args(OllamaWorkloadType))
 
 # =============================================================================
 # Tenants
@@ -283,6 +296,7 @@ async def list_callback_replies(
         "limit": per_page,
         "offset": (page - 1) * per_page,
     }
+
     if project_id:
         where_clauses.append("m.project_id = :project_id")
         params["project_id"] = project_id
@@ -395,6 +409,150 @@ async def list_callback_replies(
     }
 
 
+async def get_ai_route_status(
+    workload_type: str | None = None,
+) -> dict[str, Any]:
+    """回傳目前 AI/Ollama provider routing 的只讀狀態，供 Operator Console 顯示。"""
+    workload = _validate_ai_route_workload(workload_type)
+    policy_order = _ai_route_policy_order(workload)
+    checked_at = _utc_now_naive()
+
+    try:
+        route = await get_ollama_failover_manager().select_provider(task_type=workload)
+    except Exception as exc:
+        return {
+            "schema_version": _AI_ROUTE_STATUS_SCHEMA_VERSION,
+            "workload_type": workload,
+            "policy_order": policy_order,
+            "selected_provider": None,
+            "selected_url": None,
+            "selected_model": None,
+            "fallback_chain": [],
+            "route_reason": "route_check_failed",
+            "route_source": "ollama_failover_manager",
+            "route_error": str(exc),
+            "health": {},
+            "checked_at": checked_at,
+        }
+
+    return {
+        "schema_version": _AI_ROUTE_STATUS_SCHEMA_VERSION,
+        "workload_type": workload,
+        "policy_order": policy_order,
+        "selected_provider": route.primary.provider_name,
+        "selected_url": route.primary.url or None,
+        "selected_model": route.primary.model,
+        "fallback_chain": [
+            _ai_route_runtime_endpoint_item(endpoint, priority=index + 2)
+            for index, endpoint in enumerate(route.fallback_chain)
+        ],
+        "route_reason": route.routing_reason,
+        "route_source": "ollama_failover_manager",
+        "route_error": None,
+        "health": _ai_route_health_map(route),
+        "checked_at": checked_at,
+    }
+
+
+def _validate_ai_route_workload(workload_type: str | None) -> OllamaWorkloadType:
+    """Normalize and validate workload filter for the public route status endpoint."""
+    workload = str(workload_type or "deep_rca").strip() or "deep_rca"
+    if workload not in _AI_ROUTE_WORKLOADS:
+        allowed = ", ".join(sorted(_AI_ROUTE_WORKLOADS))
+        raise HTTPException(
+            status_code=status.HTTP_422_UNPROCESSABLE_CONTENT,
+            detail=f"Unsupported workload_type: {workload}. Allowed: {allowed}",
+        )
+    return workload  # type: ignore[return-value]
+
+
+def _ai_route_policy_order(workload: OllamaWorkloadType) -> list[dict[str, Any]]:
+    """Expose configured policy order: GCP-A -> GCP-B -> 111 -> Gemini."""
+    items = [
+        _ai_route_policy_endpoint_item(endpoint, priority=index + 1)
+        for index, endpoint in enumerate(resolve_ollama_order(workload))
+    ]
+    items.append({
+        "priority": len(items) + 1,
+        "provider_name": "gemini",
+        "url": None,
+        "workload_type": workload,
+        "reason": "final_cloud_fallback_after_all_ollama_endpoints",
+        "role": "final_fallback",
+        "runtime": "cloud",
+    })
+    return items
+
+
+def _ai_route_policy_endpoint_item(
+    endpoint: OllamaEndpointSelection,
+    *,
+    priority: int,
+) -> dict[str, Any]:
+    role = {
+        "ollama_gcp_a": "primary",
+        "ollama_gcp_b": "secondary",
+        "ollama_local": "local_fallback",
+    }.get(endpoint.provider_name, "ollama")
+    return {
+        "priority": priority,
+        "provider_name": endpoint.provider_name,
+        "url": endpoint.url,
+        "workload_type": endpoint.workload_type,
+        "reason": endpoint.reason,
+        "role": role,
+        "runtime": "ollama",
+    }
+
+
+def _ai_route_runtime_endpoint_item(
+    endpoint: OllamaEndpoint,
+    *,
+    priority: int,
+) -> dict[str, Any]:
+    return {
+        "priority": priority,
+        "provider_name": endpoint.provider_name,
+        "url": endpoint.url or None,
+        "model": endpoint.model,
+        "runtime": "ollama" if endpoint.provider_name.startswith("ollama") else "cloud",
+    }
+
+
+def _ai_route_health_map(route: OllamaRoutingResult) -> dict[str, dict[str, Any]]:
+    """Convert failover health reports into provider keyed status for the UI."""
+    health: dict[str, dict[str, Any]] = {
+        "ollama_gcp_a": _ai_route_health_item(route.health_gcp_a),
+    }
+    if route.health_gcp_b:
+        health["ollama_gcp_b"] = _ai_route_health_item(route.health_gcp_b)
+    else:
+        health["ollama_gcp_b"] = _ai_route_not_checked_health_item()
+    if route.health_local:
+        health["ollama_local"] = _ai_route_health_item(route.health_local)
+    else:
+        health["ollama_local"] = _ai_route_not_checked_health_item()
+    return health
+
+
+def _ai_route_health_item(report: HealthReport) -> dict[str, Any]:
+    payload = report.to_dict()
+    payload["checked"] = True
+    return payload
+
+
+def _ai_route_not_checked_health_item() -> dict[str, Any]:
+    return {
+        "status": "not_checked",
+        "host": "",
+        "latency_ms": None,
+        "reason": "standby_not_checked_primary_healthy",
+        "checked_at": None,
+        "from_cache": False,
+        "checked": False,
+    }
+
+
 def _timeline_item(
     *,
     ts: Any,