@runtypelabs/cli 2.10.7 → 2.10.9

package/dist/index.js

@@ -57,6 +57,7 @@ var require_default_models = __commonJS({
       // OpenAI models (routed via OpenAI direct, with Vercel fallback)
       { provider: "runtype", modelId: "gpt-5.4", isDefault: false },
       { provider: "runtype", modelId: "gpt-5.4-mini", isDefault: false },
+      { provider: "runtype", modelId: "gpt-5.4-nano", isDefault: false },
       // Anthropic models (routed via Anthropic direct, with Vertex/Bedrock fallback)
       { provider: "runtype", modelId: "claude-sonnet-4-6", isDefault: false },
       { provider: "runtype", modelId: "claude-haiku-4-5", isDefault: false },
@@ -69,7 +70,13 @@ var require_default_models = __commonJS({
       { provider: "runtype", modelId: "grok-4", isDefault: false },
       { provider: "runtype", modelId: "grok-4-fast", isDefault: false },
       // Inception models (routed via Vercel)
-      { provider: "runtype", modelId: "mercury-2", isDefault: false }
+      { provider: "runtype", modelId: "mercury-2", isDefault: false },
+      // Moonshot models (routed via Cloudflare Workers AI)
+      { provider: "runtype", modelId: "kimi-k2.5", isDefault: false },
+      // NVIDIA models (routed via Cloudflare Workers AI)
+      { provider: "runtype", modelId: "nemotron-3-120b-a12b", isDefault: false },
+      // Google models (routed via Cloudflare Workers AI)
+      { provider: "runtype", modelId: "gemma-4-26b-a4b-it", isDefault: false }
     ];
     exports.DEFAULT_MODEL_ID = (_b = (_a = exports.DEFAULT_MODELS_FOR_NEW_ACCOUNTS.find((m) => m.isDefault)) === null || _a === void 0 ? void 0 : _a.modelId) !== null && _b !== void 0 ? _b : "qwen/qwen3.5-9b";
     exports.FALLBACK_MODEL_ID = "gpt-5.4-nano";
@@ -14448,7 +14455,7 @@ var require_provider_routing = __commonJS({
   "../shared/dist/provider-routing.js"(exports) {
     "use strict";
     Object.defineProperty(exports, "__esModule", { value: true });
-    exports.GATEWAY_PROVIDERS = exports.INFERENCE_PROVIDERS = exports.DIRECT_PROVIDERS = exports.ROUTED_MODEL_DISPLAY_NAMES = exports.BASE_MODEL_PROVIDER_MAP = exports.PROVIDER_ROUTING_CONFIG = exports.PROVIDER_API_KEY_MAP = exports.PLATFORM_KEY_PROVIDERS = void 0;
+    exports.GATEWAY_PROVIDERS = exports.INFERENCE_PROVIDERS = exports.DIRECT_PROVIDERS = exports.ROUTED_MODEL_DISPLAY_NAMES = exports.BASE_MODEL_PROVIDER_MAP = exports.PROVIDER_ROUTING_CONFIG = exports.CLOUDFLARE_GATEWAY_ALLOWLIST = exports.PROVIDER_API_KEY_MAP = exports.PLATFORM_KEY_PROVIDERS = void 0;
     exports.isRoutedModel = isRoutedModel;
     exports.isDirectAccess = isDirectAccess;
     exports.isGatewayAccess = isGatewayAccess;
@@ -14470,8 +14477,10 @@ var require_provider_routing = __commonJS({
     exports.inferModelCreator = inferModelCreator2;
     exports.getDefaultProvider = getDefaultProvider;
     exports.isVercelSupported = isVercelSupported;
+    exports.isCloudflareSupported = isCloudflareSupported;
     exports.generateVercelModelId = generateVercelModelId;
     exports.getDefaultRoutingConfig = getDefaultRoutingConfig;
+    exports.generateCloudflareModelId = generateCloudflareModelId;
     exports.getDefaultProviderMapping = getDefaultProviderMapping;
     exports.getRoutingConfigWithDefaults = getRoutingConfigWithDefaults;
     exports.getProviderModelIdWithDefaults = getProviderModelIdWithDefaults;
@@ -14482,7 +14491,8 @@ var require_provider_routing = __commonJS({
       "anthropic",
       "google",
       "xai",
-      "vercel"
+      "vercel",
+      "cloudflare"
     ]);
     exports.PROVIDER_API_KEY_MAP = {
       "openai": "openaiKey",
@@ -14491,6 +14501,7 @@ var require_provider_routing = __commonJS({
       "xai": "xaiKey",
       "mixlayer": "modelsocketKey",
       "vercel": "vercelGatewayKey",
+      "cloudflare": "cloudflareGatewayToken",
       "togetherai": "togetheraiKey",
       "bedrock": "bedrockAccessKey",
       // Also needs bedrockSecretKey
@@ -14507,6 +14518,9 @@ var require_provider_routing = __commonJS({
       "mock": null
       // Dev-only, no API key needed
     };
+    exports.CLOUDFLARE_GATEWAY_ALLOWLIST = /* @__PURE__ */ new Set([
+      "gpt-5.4-nano"
+    ]);
     exports.PROVIDER_ROUTING_CONFIG = {
       // Claude 4.5 Sonnet - Route through Vercel AI Gateway
       "claude-sonnet-4-5": {
@@ -14556,6 +14570,20 @@ var require_provider_routing = __commonJS({
         ],
         fallback: ["openai"]
       },
+      // GPT-5.3 Codex - completion model, route through Vercel + OpenAI fallback
+      "gpt-5.3-codex": {
+        providers: [
+          { provider: "vercel", weight: 100 }
+        ],
+        fallback: ["openai"]
+      },
+      // GPT-5.4 Nano - CF Gateway primary for testing, Vercel + OpenAI as fallback
+      "gpt-5.4-nano": {
+        providers: [
+          { provider: "cloudflare", weight: 100 }
+        ],
+        fallback: ["vercel", "openai"]
+      },
       // Gemini 2.5 Pro - Route through Vercel AI Gateway
       "gemini-2.5-pro": {
         providers: [
@@ -14653,6 +14681,26 @@ var require_provider_routing = __commonJS({
           { provider: "vercel", weight: 100 }
         ],
         fallback: ["openai"]
+      },
+      // Kimi K2.5 - Route through Cloudflare Workers AI
+      "kimi-k2.5": {
+        providers: [
+          { provider: "workers-ai", weight: 100 }
+        ],
+        fallback: ["vercel"]
+      },
+      // NVIDIA Nemotron-3 120B - Route through Cloudflare Workers AI (no Vercel equivalent)
+      "nemotron-3-120b-a12b": {
+        providers: [
+          { provider: "workers-ai", weight: 100 }
+        ]
+      },
+      // Google Gemma 4 26B - Route through Cloudflare Workers AI
+      "gemma-4-26b-a4b-it": {
+        providers: [
+          { provider: "workers-ai", weight: 100 }
+        ],
+        fallback: ["vercel"]
       }
     };
     var MANUAL_PROVIDER_MAP_OVERRIDES = {
@@ -14677,6 +14725,15 @@ var require_provider_routing = __commonJS({
       },
       "deepseek-v3": {
         "togetherai": "togetherai/deepseek-ai/DeepSeek-V3"
+      },
+      "kimi-k2.5": {
+        "workers-ai": "@cf/moonshotai/kimi-k2.5"
+      },
+      "nemotron-3-120b-a12b": {
+        "workers-ai": "@cf/nvidia/nemotron-3-120b-a12b"
+      },
+      "gemma-4-26b-a4b-it": {
+        "workers-ai": "@cf/google/gemma-4-26b-a4b-it"
       }
     };
     exports.BASE_MODEL_PROVIDER_MAP = (() => {
@@ -14802,10 +14859,12 @@ var require_provider_routing = __commonJS({
       "vertex-anthropic",
       "tinfoil",
       "generic-openai",
-      "mock"
+      "mock",
+      "workers-ai"
     ]);
     exports.GATEWAY_PROVIDERS = /* @__PURE__ */ new Set([
       "vercel",
+      "cloudflare",
       "openrouter",
       "azure"
     ]);
@@ -15145,6 +15204,9 @@ var require_provider_routing = __commonJS({
       const creator = inferModelCreator2(modelId);
       return !NON_VERCEL_CREATORS.has(creator);
     }
+    function isCloudflareSupported(modelId) {
+      return exports.CLOUDFLARE_GATEWAY_ALLOWLIST.has(modelId);
+    }
     var CREATOR_TO_VERCEL_PROVIDER = {
       "openai": "openai",
       "anthropic": "anthropic",
@@ -15177,18 +15239,33 @@ var require_provider_routing = __commonJS({
       const config2 = {
         providers: [{ provider: "vercel", weight: 100 }]
       };
+      const fallbacks = [];
+      if (isCloudflareSupported(baseModel)) {
+        fallbacks.push("cloudflare");
+      }
       const fallbackProvider = getFallbackProvider(creator);
       if (fallbackProvider) {
-        config2.fallback = [fallbackProvider];
+        fallbacks.push(fallbackProvider);
+      }
+      if (fallbacks.length > 0) {
+        config2.fallback = fallbacks;
       }
       return config2;
     }
+    function generateCloudflareModelId(modelId) {
+      const creator = inferModelCreator2(modelId);
+      const cfProvider = CREATOR_TO_VERCEL_PROVIDER[creator] || creator;
+      return `${cfProvider}/${modelId}`;
+    }
     function getDefaultProviderMapping(baseModel) {
       const creator = inferModelCreator2(baseModel);
       const mapping = {};
       if (isVercelSupported(baseModel)) {
         mapping["vercel"] = generateVercelModelId(baseModel);
       }
+      if (isCloudflareSupported(baseModel)) {
+        mapping["cloudflare"] = generateCloudflareModelId(baseModel);
+      }
       const fallbackProvider = getFallbackProvider(creator);
       if (fallbackProvider) {
         mapping[fallbackProvider] = baseModel;
@@ -26620,16 +26697,17 @@ function createSandboxInstructions(provider) {
       "--- Deploy with Preview (Cloudflare Sandbox) ---",
       "You also have `deploy_sandbox` to deploy a persistent web server and get a live preview URL.",
       "Call shape:",
-      '{ "code": "...", "files": { "public/index.html": "..." }, "packageJson": "{\\"dependencies\\": {\\"hono\\": \\"^4.0.0\\"}}", "language": "typescript", "port": 3000 }',
+      '{ "code": "...", "files": { "public/index.html": "..." }, "packageJson": "{\\"dependencies\\": {\\"hono\\": \\"^4.0.0\\"}}", "language": "typescript", "port": 8080 }',
       "Deploy rules:",
       "1. Use this when you need to run a web server (Hono, Express, FastAPI, etc.) that the user can visit in their browser.",
       "2. The `code` is the main entry point file. The `packageJson` string provides npm dependencies.",
-      "3. The `files` parameter writes additional files (path \u2192 content). Keys must be plain file paths.",
-      "4. Bind the server to the specified `port` (default 3000).",
+      '3. The `files` parameter writes additional files (path \u2192 content). Keys must be plain relative file paths. Parent directories are created automatically, so `{ "public/index.html": "..." }` just works.',
+      "4. Bind the server to the specified `port` (default 8080). Port 3000 is reserved by the sandbox API control plane \u2014 pick 8080, 8000, 5173, or any other non-3000 port.",
       "5. Each sandbox is a persistent container with snapshot-backed warm starts.",
       "6. The returned `previewUrl` is a REAL live URL the user can open in their browser.",
       "7. The sandbox persists for 2 hours after the last activity, then auto-cleans up (state survives via snapshot).",
-      "8. Prefer this over `run_sandbox_code` when building web apps, APIs, or anything with a UI."
+      "8. Prefer this over `run_sandbox_code` when building web apps, APIs, or anything with a UI.",
+      '9. On failure the response includes a `stage` field (e.g. "npm_install", "start_process") and captured stdout/stderr in `output` \u2014 use these to diagnose without re-running.'
     ].join("\n");
   }
   return [
@@ -26885,7 +26963,7 @@ function createDeployCfSandboxLocalTool(client, debugMode) {
         },
         port: {
           type: "number",
-          description: "Port the server listens on (default: 3000)"
+          description: "Port the server listens on (default: 8080). Port 3000 is reserved by the sandbox API control plane and will be rejected \u2014 use 8080, 8000, 5173, or any other non-3000 port."
         },
         startCommand: {
           type: "string",
@@ -26893,7 +26971,7 @@ function createDeployCfSandboxLocalTool(client, debugMode) {
         },
         files: {
           type: "object",
-          description: 'Additional files to write (path \u2192 content), e.g. { "public/index.html": "<html>..." }. Keys must be plain file paths. Use this for multi-file projects.',
+          description: 'Additional files to write (path \u2192 content), e.g. { "public/index.html": "<html>..." }. Keys must be plain relative file paths; parent directories are created automatically. Use this for multi-file projects.',
           additionalProperties: { type: "string" }
         }
       },

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@runtypelabs/cli",
-  "version": "2.10.7",
+  "version": "2.10.9",
   "description": "Command-line interface for Runtype AI platform",
   "type": "module",
   "main": "dist/index.js",
@@ -23,8 +23,8 @@
     "micromatch": "^4.0.8",
     "yaml": "^2.8.3",
     "@runtypelabs/ink-components": "0.3.1",
-    "@runtypelabs/terminal-animations": "0.2.0",
-    "@runtypelabs/sdk": "1.15.2"
+    "@runtypelabs/sdk": "1.16.0",
+    "@runtypelabs/terminal-animations": "0.2.0"
   },
   "devDependencies": {
     "@types/micromatch": "^4.0.9",
@@ -37,7 +37,7 @@
     "tsx": "^4.7.1",
     "typescript": "^5.3.3",
     "vitest": "^4.0.18",
-    "@runtypelabs/shared": "1.4.2"
+    "@runtypelabs/shared": "1.4.3"
   },
   "engines": {
     "node": ">=18.0.0"