feat(model): add OrcaRouter as a named model provider

backend/consts/provider.py

@@ -8,6 +8,7 @@ class ProviderEnum(str, Enum):
     MODELENGINE = "modelengine"
     DASHSCOPE = "dashscope"
     TOKENPONY = "tokenpony"
+    ORCAROUTER = "orcarouter"
 
 
 # Silicon Flow
@@ -26,3 +27,7 @@ class ProviderEnum(str, Enum):
 
 # ModelEngine
 # Base URL and API key are loaded from environment variables at runtime
+
+# OrcaRouter
+ORCAROUTER_BASE_URL = "https://api.orcarouter.ai/v1/"
+ORCAROUTER_GET_URL = "https://api.orcarouter.ai/v1/models"

backend/services/model_capacity_suggestion_service.py

@@ -128,6 +128,7 @@ class CapacitySuggestionResult:
     ("deepseek", "deepseek"),
     ("jina", "jina"),
     ("tokenpony", "tokenpony"),
+    ("orcarouter", "orcarouter"),
     ("bytedance", "volcengine"),
 )
 

backend/services/model_gateway_service.py

@@ -38,6 +38,7 @@
     "siliconflow": "siliconflow",
     "openai": "openai",
     "tokenpony": "tokenpony",
+    "orcarouter": "orcarouter",
     "jina": "jina",
     "cohere": "cohere",
     "modelengine": "modelengine",

backend/services/model_management_service.py

@@ -16,6 +16,7 @@
     DASHSCOPE_BASE_URL,
     DASHSCOPE_REALTIME_BASE_URL,
     TOKENPONY_BASE_URL,
+    ORCAROUTER_BASE_URL,
 )
 
 from database.model_management_db import (
@@ -418,6 +419,8 @@ async def batch_create_models_for_tenant(user_id: str, tenant_id: str, batch_pay
             model_url = DASHSCOPE_REALTIME_BASE_URL if model_type in ("stt", "tts") else DASHSCOPE_BASE_URL
         elif provider == ProviderEnum.TOKENPONY.value:
             model_url = TOKENPONY_BASE_URL
+        elif provider == ProviderEnum.ORCAROUTER.value:
+            model_url = ORCAROUTER_BASE_URL
         else:
             model_url = ""
 

backend/services/model_provider_service.py

@@ -14,6 +14,7 @@
 from services.providers.tokenpony_provider import TokenPonyModelProvider
 from services.providers.dashscope_provider import DashScopeModelProvider
 from services.providers.modelengine_provider import ModelEngineProvider, get_model_engine_raw_url, MODEL_ENGINE_NORTH_PREFIX
+from services.providers.orcarouter_provider import OrcaRouterModelProvider
 from utils.model_name_utils import split_repo_name, add_repo_to_name
 
 logger = logging.getLogger("model_provider")
@@ -48,6 +49,9 @@ async def get_provider_models(model_data: dict) -> List[dict]:
     elif model_data["provider"] == ProviderEnum.TOKENPONY.value:
         provider = TokenPonyModelProvider()
         model_list = await provider.get_models(model_data)
+    elif model_data["provider"] == ProviderEnum.ORCAROUTER.value:
+        provider = OrcaRouterModelProvider()
+        model_list = await provider.get_models(model_data)
 
     return model_list
 

backend/services/providers/__init__.py

@@ -2,10 +2,12 @@
 from services.providers.base import AbstractModelProvider
 from services.providers.silicon_provider import SiliconModelProvider
 from services.providers.modelengine_provider import ModelEngineProvider, get_model_engine_raw_url
+from services.providers.orcarouter_provider import OrcaRouterModelProvider
 
 __all__ = [
     "AbstractModelProvider",
     "SiliconModelProvider",
     "ModelEngineProvider",
+    "OrcaRouterModelProvider",
     "get_model_engine_raw_url",
 ]

backend/services/providers/orcarouter_provider.py

@@ -0,0 +1,80 @@
+from typing import Dict, List
+
+import httpx
+from consts.const import DEFAULT_LLM_MAX_TOKENS
+from consts.provider import ORCAROUTER_GET_URL
+from services.providers.base import (
+    AbstractModelProvider,
+    _classify_provider_error,
+    _extract_capacity_hints_from_raw,
+)
+
+
+# OrcaRouter is a chat-only gateway. The /v1/models catalog only carries
+# models that support chat (the gateway routes to upstream LLMs on demand).
+# Model IDs all use the ``orcarouter/`` namespace prefix (e.g.
+# ``orcarouter/auto``, ``orcarouter/fusion``).
+ORCAROUTER_CHAT_TYPES = ("llm", "vlm")
+
+
+def _extract_capacity_hints(raw: Dict) -> Dict:
+    return _extract_capacity_hints_from_raw(raw)
+
+
+class OrcaRouterModelProvider(AbstractModelProvider):
+    """Concrete implementation for the OrcaRouter model gateway.
+
+    OrcaRouter is an OpenAI-compatible gateway: the chat model catalog is
+    fetched from ``GET /v1/models`` and every model routes through the
+    gateway's smart router (default model ``orcarouter/auto``).
+    """
+
+    async def get_models(self, provider_config: Dict) -> List[Dict]:
+        """
+        Fetch chat models from the OrcaRouter gateway API.
+
+        Args:
+            provider_config: Configuration dict containing model_type and api_key
+
+        Returns:
+            List of chat models with canonical fields. Returns error dict if
+            the API call fails.
+        """
+        try:
+            model_type: str = provider_config["model_type"]
+            model_api_key: str = provider_config["api_key"]
+
+            # OrcaRouter only routes chat models; non-chat modalities have no
+            # gateway endpoints (e.g. /v1/embeddings returns 400).
+            if model_type not in ORCAROUTER_CHAT_TYPES:
+                return []
+
+            headers = {"Authorization": f"Bearer {model_api_key}"}
+
+            async with httpx.AsyncClient() as client:
+                response = await client.get(ORCAROUTER_GET_URL, headers=headers)
+                response.raise_for_status()
+                # OpenAI-standard response: model list under the "data" array
+                all_models: List[Dict] = response.json().get("data", [])
+
+            models = []
+            for model_obj in all_models:
+                model_id = model_obj.get("id", "")
+                cleaned_model = {
+                    "id": model_id,
+                    "model_tag": "chat",
+                    "model_type": model_type,
+                    "max_tokens": DEFAULT_LLM_MAX_TOKENS,
+                }
+                cleaned_model.update(_extract_capacity_hints(model_obj))
+                models.append(cleaned_model)
+
+            return models
+        except (httpx.HTTPStatusError, httpx.ConnectTimeout, httpx.ConnectError, Exception) as e:
+            status_code = e.response.status_code if isinstance(e, httpx.HTTPStatusError) and getattr(e, "response", None) else None
+            return _classify_provider_error(
+                "OrcaRouter",
+                status_code=status_code,
+                error_message=str(e),
+                exception=e,
+            )

doc/docs/en/user-guide/agent-development/model-configuration.md

@@ -163,6 +163,7 @@ Nexent supports any **OpenAI-compatible** provider, including:
 - [OpenAI](https://platform.openai.com/)
 - [Anthropic](https://console.anthropic.com/)
 - [Moonshot](https://platform.moonshot.cn/)
+- [OrcaRouter](https://www.orcarouter.ai)
 
 Getting started:
 
@@ -171,6 +172,8 @@ Getting started:
 3. Locate the API endpoint (usually ending with `/v1`).
 4. Click **Add Custom Model** in Nexent and fill in the required fields.
 
+> **Batch-add OrcaRouter**: in the batch-add dialog, select **OrcaRouter** as the model provider, enter your `sk-orca-…` API key, and click **Fetch Models**. The gateway model catalog is loaded automatically from `https://api.orcarouter.ai/v1/models` (e.g. `orcarouter/auto` for smart routing). OrcaRouter also runs gateway-level, zero-trust security for AI agents on the same endpoint — screening every prompt/response and governing every tool call on a default-deny basis, with no application code changes.
+
 #### Multimodal Models
 
 Use the same API key and URL as LLMs but specify a multimodal model name, for example **Qwen/Qwen2.5-VL-32B-Instruct** on SiliconFlow.

doc/docs/zh/user-guide/agent-development/model-configuration.md

@@ -181,6 +181,7 @@ Nexent 支持任何 **遵循OpenAI API规范** 的大语言模型供应商，包
 - [OpenAI](https://platform.openai.com/)
 - [Anthropic](https://console.anthropic.com/)
 - [月之暗面](https://platform.moonshot.cn/)
+- [OrcaRouter](https://www.orcarouter.ai)
 
 可参考以下步骤进行模型接入：
 
@@ -189,6 +190,8 @@ Nexent 支持任何 **遵循OpenAI API规范** 的大语言模型供应商，包
 3. 在文档中查看API端点（即模型URL，一般以`/v1`为结尾）；
 4. 在Nexent模型配置页面点击添加自定义模型，填入必备信息，即可接入。
 
+> **批量接入 OrcaRouter**：在批量添加对话框中，将模型提供商选为 **OrcaRouter**，输入 `sk-orca-…` 开头的 API Key，点击「获取模型」即可自动从 `https://api.orcarouter.ai/v1/models` 拉取网关模型目录（例如智能路由模型 `orcarouter/auto`）。OrcaRouter 还在同一端点提供面向 AI Agent 的网关级零信任安全防护——以默认拒绝（default-deny）的方式审查每次提示词/响应并管控每次工具调用，无需修改任何应用代码。
+
 #### 🎭 多模态模型
 
 使用与大语言模型相同的API Key和模型URL，但指定多模态模型名称，如硅基流动提供的**Qwen/Qwen2.5-VL-32B-Instruct**。

frontend/app/[locale]/models/components/model/ModelAddDialog.tsx

@@ -33,6 +33,7 @@
 import { useSiliconModelList } from "@/hooks/model/useSiliconModelList";
 import { useDashscopeModelList } from "@/hooks/model/useDashscopeModelList";
 import { useTokenPonyModelList } from "@/hooks/model/useTokenponyModelList";
+import { useOrcaRouterModelList } from "@/hooks/model/useOrcaRouterModelList";
 import log from "@/lib/logger";
 import { publicAsset } from "@/lib/publicAsset";
 import {
@@ -375,6 +376,14 @@
     setLoadingModelList,
     tenantId,
   });
+  const orcarouterHook = useOrcaRouterModelList({
+    form,
+    setModelList,
+    setSelectedModelIds,
+    setShowModelList,
+    setLoadingModelList,
+    tenantId,
+  });
   let getModelList;
   let getProviderSelectedModalList;
 
@@ -385,6 +394,8 @@
     ({ getModelList, getProviderSelectedModalList } = dashscopeHook);
   } else if (form.provider === "tokenpony") {
     ({ getModelList, getProviderSelectedModalList } = tokenponyHook);
+  } else if (form.provider === "orcarouter") {
+    ({ getModelList, getProviderSelectedModalList } = orcarouterHook);
   }
   // Reset form to default state
   const resetForm = useCallback(() => {
@@ -1441,6 +1452,9 @@
               <Option value="silicon">{t("model.provider.silicon")}</Option>
               <Option value="dashscope">{t("model.provider.dashscope")}</Option>
               <Option value="tokenpony">{t("model.provider.tokenpony")}</Option>
+              <Option value="orcarouter">
+                {t("model.provider.orcarouter")}
+              </Option>
             </Select>
             {/* ModelEngine URL input (only when provider is ModelEngine) */}
             {form.provider === "modelengine" && (

frontend/app/[locale]/models/components/model/ModelDeleteDialog.tsx

@@ -219,6 +219,8 @@ export const ModelDeleteDialog = ({
         return t("model.source.dashscope");
       case MODEL_SOURCES.TOKENPONY:
         return t("model.source.tokenpony");
+      case MODEL_SOURCES.ORCAROUTER:
+        return t("model.source.orcarouter");
       case MODEL_SOURCES.VOLCENGINE:
         return t("model.provider.volcengine");
       default:
@@ -267,6 +269,12 @@ export const ModelDeleteDialog = ({
           text: "text-cyan-600",
           border: "border-cyan-100",
         };
+      case MODEL_SOURCES.ORCAROUTER:
+        return {
+          bg: "bg-sky-50",
+          text: "text-sky-600",
+          border: "border-sky-100",
+        };
       case MODEL_SOURCES.VOLCENGINE:
         return {
           bg: "bg-pink-50",
@@ -313,6 +321,14 @@ export const ModelDeleteDialog = ({
         return <img src={publicAsset("/aliyuncs.png")} alt="DashScope" className="w-5 h-5" />;
       case MODEL_SOURCES.TOKENPONY:
         return <img src={publicAsset("/tokenpony.png")} alt="TokenPony" className="w-5 h-5" />;
+      case MODEL_SOURCES.ORCAROUTER:
+        return (
+          <img
+            src={publicAsset("/orcarouter.svg")}
+            alt="OrcaRouter"
+            className="w-5 h-5"
+          />
+        );
       case MODEL_SOURCES.VOLCENGINE:
         return (
           <img src={publicAsset("/volcengine.png")} alt="VolcEngine" className="w-5 h-5" />

frontend/app/[locale]/models/components/model/ModelListCard.tsx

@@ -174,6 +174,8 @@
       return t("model.source.dashscope");
     } else if (model.source === "tokenpony") {
       return t("model.source.tokenpony");
+    } else if (model.source === "orcarouter") {
+      return t("model.source.orcarouter");
     } else if (model.source === "volcengine") {
       return t("model.provider.volcengine");
     } else if (model.source === "OpenAI-API-Compatible") {
@@ -191,6 +193,7 @@
     silicon: filteredModels.filter((m) => m.source === "silicon"),
     dashscope: filteredModels.filter((m) => m.source === "dashscope"),
     tokenpony: filteredModels.filter((m) => m.source === "tokenpony"),
+    orcarouter: filteredModels.filter((m) => m.source === "orcarouter"),
     volcengine: filteredModels.filter((m) => m.source === "volcengine"),
     custom: filteredModels.filter((m) => m.source === "OpenAI-API-Compatible"),
   };
@@ -448,6 +451,54 @@
             ))}
           </Select.OptGroup>
         )}
+        {groupedModels.orcarouter.length > 0 && (
+          <Select.OptGroup label={t("model.group.orcarouter")}>
+            {groupedModels.orcarouter.map((model) => (
+              <Option
+                key={`${type}-${model.displayName}-orcarouter`}
+                value={model.displayName}
+              >
+                <div
+                  className="flex items-center justify-between"
+                  style={{ minWidth: 0 }}
+                >
+                  <div
+                    className="flex items-center font-medium truncate"
+                    style={{ flex: "1 1 auto", minWidth: 0 }}
+                    title={model.displayName}
+                  >
+                    <img
+                      src={getProviderIconByUrl(model.apiUrl)}
+                      alt="provider"
+                      className="w-4 h-4 rounded mr-2 flex-shrink-0"
+                    />
+                    <span className="truncate">{model.displayName}</span>
+                  </div>
+                  <div
+                    style={{
+                      flex: "0 0 auto",
+                      display: "flex",
+                      alignItems: "center",
+                      marginLeft: "8px",
+                    }}
+                  >
+                    <Tooltip title={t("model.status.tooltip")}>
+                      <span
+                        onClick={(e) => handleStatusClick(e, model.displayName)}
+                        onMouseDown={(e: React.MouseEvent) => {
+                          e.stopPropagation();
+                          e.preventDefault();
+                        }}
+                        style={getStatusStyle(model.connect_status)}
+                        className="status-indicator"
+                      />
+                    </Tooltip>
+                  </div>
+                </div>
+              </Option>
+            ))}
+          </Select.OptGroup>
+        )}
         {groupedModels.volcengine.length > 0 && (
           <Select.OptGroup label={t("model.group.volcengine")}>
             {groupedModels.volcengine.map((model) => (

frontend/const/modelConfig.ts

@@ -23,6 +23,7 @@ export const MODEL_SOURCES = {
   DASHSCOPE: "dashscope",
   TOKENPONY: "tokenpony",
   VOLCENGINE: "volcengine",
+  ORCAROUTER: "orcarouter",
 } as const;
 
 // Model status constants
@@ -50,6 +51,7 @@ export const MODEL_PROVIDER_KEYS = [
   "tokenpony",
   "dashscope",
   "volcengine",
+  "orcarouter",
 ] as const;
 
 export type ModelProviderKey = (typeof MODEL_PROVIDER_KEYS)[number];
@@ -65,6 +67,7 @@ export const PROVIDER_HINTS: Record<ModelProviderKey, string> = {
   tokenpony: "tokenpony",
   dashscope: "dashscope",
   volcengine: "bytedance",
+  orcarouter: "orcarouter",
 };
 
 // Icon filenames for providers
@@ -78,6 +81,7 @@ export const PROVIDER_ICON_MAP: Record<ModelProviderKey, string> = {
   dashscope: publicAsset("/aliyuncs.png"),
   tokenpony: publicAsset("/tokenpony.png"),
   volcengine: publicAsset("/volcengine.png"),
+  orcarouter: publicAsset("/orcarouter.svg"),
 };
 
 export const OFFICIAL_PROVIDER_ICON = publicAsset("/modelengine-logo.png");
@@ -96,6 +100,7 @@ export const PROVIDER_LINKS: Record<string, string> = {
   dashscope: "https://dashscope.aliyun.com/",
   tokenpony: "https://www.tokenpony.cn/",
   volcengine: "https://www.volcengine.com/",
+  orcarouter: "https://www.orcarouter.ai",
 };
 
 // User role constants