feat(llm): add OrcaRouter as a first-class model provider

package.json

@@ -8,7 +8,7 @@
     "test": "pnpm -r --if-present run test",
 
     "build:base": "pnpm --filter scorpio.di run build && pnpm --filter scorpio.command run build && pnpm --filter scorpio.ai run build && pnpm --filter scorpio.tools.filesystem run build && pnpm --filter scorpio.tools.command run build",
-    "build:llm": "node scripts/fetch-models-dev.js && pnpm --filter scorpio.llm run build && pnpm --filter scorpio.llm.openai run build && pnpm --filter scorpio.llm.anthropic run build && pnpm --filter scorpio.llm.gemini run build && pnpm --filter scorpio.llm.ollama run build && pnpm --filter scorpio.llm.cohere run build && pnpm --filter scorpio.llm.voyage run build",
+    "build:llm": "node scripts/fetch-models-dev.js && pnpm --filter scorpio.llm run build && pnpm --filter scorpio.llm.openai run build && pnpm --filter scorpio.llm.anthropic run build && pnpm --filter scorpio.llm.gemini run build && pnpm --filter scorpio.llm.ollama run build && pnpm --filter scorpio.llm.cohere run build && pnpm --filter scorpio.llm.voyage run build && pnpm --filter scorpio.llm.orcarouter run build",
     "build:saver": "pnpm --filter scorpio.saver run build && pnpm --filter scorpio.saver.file run build && pnpm --filter scorpio.saver.sqlite run build && pnpm --filter scorpio.saver.postgres run build",
     "build:schema": "pnpm --filter plugin.schema run build && pnpm --filter @sbot/shared run build",
     "build:agents": "pnpm --filter agent.skill run build && pnpm --filter agent.agenda run build && pnpm --filter agent.memory run build && pnpm --filter agent.note run build && pnpm --filter agent.wiki run build",

packages/apps/docs/guide/models.md

@@ -14,6 +14,7 @@ Fill in provider, API key, base URL, and model name. The same connection can be
 | Google Gemini | Gemini 2.0 / 2.5 Pro, Flash |
 | Gemini Image | Gemini image-generation models |
 | Ollama | Local models via Ollama runtime |
+| OrcaRouter | OpenAI-compatible gateway for models and agents (adaptive routing, failover, zero-markup inference) |
 | OpenAI-compatible | Azure OpenAI, Groq, Mistral, DeepSeek, Qwen, Together, etc. |
 
 Any endpoint that implements the OpenAI chat completions API can be used as a provider — pick **OpenAI-compatible** and override the base URL.

packages/apps/docs/zh/guide/models.md

@@ -14,6 +14,7 @@
 | Google Gemini | Gemini 2.0 / 2.5 Pro、Flash |
 | Gemini Image | Gemini 图像生成模型 |
 | Ollama | 通过 Ollama 运行时使用本地模型 |
+| OrcaRouter | 面向模型与 Agent 的 OpenAI 兼容网关（自适应路由、故障转移、零加成推理） |
 | OpenAI 兼容 | Azure OpenAI、Groq、Mistral、DeepSeek、Qwen、Together 等 |
 
 任何实现了 OpenAI Chat Completions 接口的服务都可以作为供应商使用 —— 选择 **OpenAI 兼容** 并覆盖 Base URL 即可。

packages/backend/llm/scorpio.llm.orcarouter/.gitignore

@@ -0,0 +1,4 @@
+/node_modules
+/dist
+/package-lock.json
+/tsconfig.tsbuildinfo

packages/backend/llm/scorpio.llm.orcarouter/package.json

@@ -0,0 +1,20 @@
+{
+  "name": "scorpio.llm.orcarouter",
+  "version": "0.0.1",
+  "description": "OrcaRouter model provider for scorpio.llm",
+  "main": "dist/index.js",
+  "types": "dist/index.d.ts",
+  "scripts": { "build": "rimraf dist && tsc -b --force" },
+  "dependencies": {
+    "@langchain/core": "catalog:",
+    "@langchain/openai": "^1.4.7",
+    "scorpio.llm": "workspace:*",
+    "scorpio.llm.openai": "workspace:*"
+  },
+  "private": true,
+  "devDependencies": {
+    "@types/node": "catalog:",
+    "rimraf": "catalog:",
+    "typescript": "catalog:"
+  }
+}

packages/backend/llm/scorpio.llm.orcarouter/src/index.ts

@@ -0,0 +1,30 @@
+import { type LlmProviderRegistry, llmProviderRegistry } from "scorpio.llm";
+import { OpenAIModelService } from "scorpio.llm.openai";
+
+/** OrcaRouter provider id, kept stable so saved model configs keep working. */
+export const OrcaRouterProvider = "orcarouter";
+
+export function registerOrcaRouterProvider(registry: LlmProviderRegistry = llmProviderRegistry): void {
+  registry.registerModel({
+    type: OrcaRouterProvider,
+    label: "OrcaRouter",
+    configSchema: {},
+    defaults: { baseURL: "https://api.orcarouter.ai/v1", model: "orcarouter/fusion" },
+    apiKeyRequired: true,
+    createModel: config => {
+      const service = new OpenAIModelService(config);
+      service.initialize();
+      return service;
+    },
+    listModels: async config => {
+      if (!config.apiKey) throw new Error("apiKey is required for OrcaRouter");
+      const base = (config.baseURL || "https://api.orcarouter.ai/v1").replace(/\/$/, "");
+      const res = await fetch(`${base}/models`, {
+        headers: { Authorization: `Bearer ${config.apiKey}` },
+      });
+      if (!res.ok) throw new Error(`Models request failed: ${res.status}`);
+      const data: any = await res.json();
+      return (data.data || []).map((model: any) => model.id as string);
+    },
+  });
+}

packages/backend/llm/scorpio.llm.orcarouter/tsconfig.json

@@ -0,0 +1,9 @@
+{
+  "extends": "../../../../tsconfig.base.json",
+  "compilerOptions": { "outDir": "dist", "rootDir": "src" },
+  "include": ["src"],
+  "references": [
+    { "path": "../scorpio.llm" },
+    { "path": "../scorpio.llm.openai" }
+  ]
+}

packages/backend/sbot/package.json

@@ -76,6 +76,7 @@
     "scorpio.llm.gemini": "workspace:*",
     "scorpio.llm.ollama": "workspace:*",
     "scorpio.llm.openai": "workspace:*",
+    "scorpio.llm.orcarouter": "workspace:*",
     "scorpio.llm.voyage": "workspace:*",
     "scorpio.saver": "workspace:*",
     "scorpio.saver.file": "workspace:*",

packages/backend/sbot/src/Core/LlmProviders.ts

@@ -3,6 +3,7 @@ import { registerCohereProvider } from "scorpio.llm.cohere";
 import { registerGeminiProvider } from "scorpio.llm.gemini";
 import { registerOllamaProvider } from "scorpio.llm.ollama";
 import { registerOpenAIProvider } from "scorpio.llm.openai";
+import { registerOrcaRouterProvider } from "scorpio.llm.orcarouter";
 import { registerVoyageProvider } from "scorpio.llm.voyage";
 
 let registered = false;
@@ -16,5 +17,6 @@ export function registerBuiltInLlmProviders(): void {
   registerOllamaProvider();
   registerCohereProvider();
   registerVoyageProvider();
+  registerOrcaRouterProvider();
   registered = true;
 }