@thanh01.pmt/curriculum-kit 1.0.8 → 1.0.10

package/dist/ai/index.cjs

@@ -175,7 +175,7 @@ function isModelAllowed(model) {
 }
 function getDesignatedFallbackChain() {
   const rawProviders = resolveApiKey("FALLBACK_AI_PROVIDER") || resolveApiKey("FALLBACK_AI_PROVIDERS") || resolveApiKey("DEFAULT_AI_PROVIDER") || "nvidia,openrouter";
-  const rawModels = resolveApiKey("FALLBACK_AI_MODEL") || resolveApiKey("FALLBACK_AI_MODELS") || resolveApiKey("DEFAULT_AI_MODEL") || "nvidia/nemotron-3-ultra-550b-a55b,@preset/coding-free,deepseek/deepseek-chat:free";
+  const rawModels = resolveApiKey("FALLBACK_AI_MODEL") || resolveApiKey("FALLBACK_AI_MODELS") || resolveApiKey("DEFAULT_AI_MODEL") || "nvidia/nemotron-3-ultra-550b-a55b:free,@preset/coding-free,deepseek/deepseek-chat:free";
   const providerList = rawProviders.split(",").map((p) => p.trim().toLowerCase()).filter((p) => p.length > 0 && isProviderEnabled(p));
   const modelList = rawModels.split(",").map((m) => m.trim()).filter((m) => m.length > 0 && isModelAllowed(m));
   const chain = [];
@@ -184,7 +184,7 @@ function getDesignatedFallbackChain() {
     if (provider === "nvidia") {
       providerModels = modelList.filter((m) => m.startsWith("nvidia/") || m.startsWith("meta/") || m.startsWith("mistralai/"));
       if (providerModels.length === 0) {
-        providerModels = ["nvidia/nemotron-3-ultra-550b-a55b", "nvidia/nemotron-3.5-lightning-30b-a3b"];
+        providerModels = ["nvidia/nemotron-3-ultra-550b-a55b:free", "nvidia/nemotron-3.5-lightning-30b-a3b"];
       }
     } else if (provider === "openrouter") {
       providerModels = modelList.filter((m) => m.startsWith("@") || m.includes("/") || m.includes(":"));
@@ -220,7 +220,7 @@ function getDesignatedFallbackChain() {
 function getDesignatedFallbackConfig() {
   const chain = getDesignatedFallbackChain();
   if (chain.length > 0) return chain[0];
-  return { provider: "nvidia", model: "nvidia/nemotron-3-ultra-550b-a55b" };
+  return { provider: "nvidia", model: "nvidia/nemotron-3-ultra-550b-a55b:free" };
 }
 function extractThoughtAndContent(rawText) {
   let text = rawText || "";
@@ -413,6 +413,32 @@ Guidelines:
 3. Keep technical terminology and official API symbols intact.
 4. Format output using clean Markdown, clear sections, bullet points, and code blocks where required.
 5. Do NOT use emojis in headings, tab titles, or UI navigation labels.`;
+  if (options.customInference) {
+    try {
+      const text = await options.customInference({
+        systemInstruction,
+        userPrompt: messages.map((m) => m.content).join("\n\n"),
+        messages,
+        temperature,
+        maxTokens,
+        onChunk: onChunk ?? (() => {
+        })
+      });
+      if (text && text.trim()) return text;
+      throw createAiInferenceError({
+        errorCode: "ERR_AI_ALL_PROVIDERS_FAILED",
+        message: "customInference returned empty output",
+        rawError: { customInference: "empty response" }
+      });
+    } catch (e) {
+      if (e?.errorCode || e?.name === "AiInferenceError") throw e;
+      throw createAiInferenceError({
+        errorCode: "ERR_AI_ALL_PROVIDERS_FAILED",
+        message: `customInference failed: ${e?.message || e}`,
+        rawError: { customInference: String(e?.message || e) }
+      });
+    }
+  }
   const formattedMessages = [
     { role: "system", content: systemInstruction },
     ...messages.map((m) => ({
@@ -425,8 +451,8 @@ Guidelines:
   const isNvidiaPreferred = requestedModel.startsWith("nvidia/") || Boolean(nvidiaKey && !openrouterKey) && isProviderEnabled("nvidia");
   if (nvidiaKey && isProviderEnabled("nvidia")) {
     const candidateNemotronModels = [
-      requestedModel.startsWith("nvidia/") ? requestedModel : "nvidia/nemotron-3-ultra-550b-a55b",
-      "nvidia/nemotron-3-ultra-550b-a55b",
+      requestedModel.startsWith("nvidia/") ? requestedModel : "nvidia/nemotron-3-ultra-550b-a55b:free",
+      "nvidia/nemotron-3-ultra-550b-a55b:free",
       "nvidia/nemotron-3.5-lightning-30b-a3b",
       "nvidia/nemotron-3-super-120b-a12b",
       "nvidia/llama-3.1-nemotron-70b-instruct"
@@ -785,7 +811,7 @@ async function runCurriculumAIInference(messages, projectContext, options = {},
 
 // src/ai/provider-factory.ts
 var DEFAULT_MODELS = {
-  nvidia: "nvidia/nemotron-3-ultra-550b-a55b",
+  nvidia: "nvidia/nemotron-3-ultra-550b-a55b:free",
   openrouter: "@preset/coding-free",
   alibaba: "qwen3.7-plus",
   dashscope: "qwen3.7-plus",
@@ -796,7 +822,7 @@ var DEFAULT_MODELS = {
 };
 var NVIDIA_MODELS = {
   codingFast: "nvidia/nemotron-3.5-lightning-30b-a3b",
-  reasoningUltra: "nvidia/nemotron-3-ultra-550b-a55b",
+  reasoningUltra: "nvidia/nemotron-3-ultra-550b-a55b:free",
   superMoE: "nvidia/nemotron-3-super-120b-a12b",
   llamaNemotron: "nvidia/llama-3.1-nemotron-70b-instruct",
   translation: "nvidia/riva-translate-4b-instruct-v2",
@@ -822,7 +848,7 @@ function getAIModel(options = {}) {
       apiKey,
       baseURL: options.baseURL || "https://integrate.api.nvidia.com/v1"
     });
-    return nvidia.chat(resolvedModelName || "nvidia/nemotron-3-ultra-550b-a55b");
+    return nvidia.chat(resolvedModelName || "nvidia/nemotron-3-ultra-550b-a55b:free");
   }
   if (provider === "openrouter") {
     const apiKey = options.apiKey || process.env.OPENROUTER_API_KEY || process.env.OPENAI_API_KEY || "";
@@ -4309,7 +4335,7 @@ var ModelPolicyResolver = class {
       {
         provider: "nvidia",
         providerLabel: "NVIDIA NIM API",
-        modelName: "nvidia/nemotron-3-ultra-550b-a55b",
+        modelName: "nvidia/nemotron-3-ultra-550b-a55b:free",
         maxTokens: 32768,
         temperature: 0.15,
         reasoningEffort: "high",
@@ -4734,7 +4760,7 @@ var PROVIDER_ENDPOINTS = {
   alibaba: "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions",
   nvidia: "https://integrate.api.nvidia.com/v1/chat/completions"
 };
-var NVIDIA_DEFAULT_MODEL = "nvidia/nemotron-3-ultra-550b-a55b";
+var NVIDIA_DEFAULT_MODEL = "nvidia/nemotron-3-ultra-550b-a55b:free";
 function resolveProviderRoute(options) {
   const openrouterKey = options.openrouterKey?.trim() || (typeof process !== "undefined" ? process.env.OPENROUTER_API_KEY || "" : "");
   const deepseekKey = options.deepseekKey?.trim() || (typeof process !== "undefined" ? process.env.DEEPSEEK_API_KEY || "" : "");