@thanh01.pmt/curriculum-kit 1.0.14 → 1.0.16

package/dist/{streamRunner-CcpmxOf1.d.cts → streamRunner-DtQq0HV_.d.cts}

@@ -15,6 +15,10 @@ interface StreamRunnerOptions {
     maxTokens?: number;
     temperature?: number;
     reasoningEffort?: 'off' | 'low' | 'medium' | 'high' | 'max';
+    /** Thời gian tối đa cho phép không nhận thêm token mới trước khi abort (mặc định 45_000ms = 45s) */
+    idleTimeoutMs?: number;
+    /** Thời gian hard cap tối đa cho toàn bộ lượt stream trước khi abort (mặc định 420_000ms = 7 phút) */
+    timeoutMs?: number;
     /**
      * Host-provided inference override (Vercel AI SDK gateway, test stubs, …).
      * When set, ALL built-in provider routes (raw-fetch NVIDIA/OpenRouter/
@@ -67,6 +71,31 @@ declare function extractThoughtAndContent(rawText: string): {
     thought: string;
     content: string;
 };
+declare const DEFAULT_ARTIFACT_STREAM_IDLE_MS = 45000;
+declare const DEFAULT_ARTIFACT_STREAM_TOTAL_MS = 420000;
+interface StreamAbortController {
+    signal: AbortSignal;
+    /** Reset the idle window (call on every received chunk). */
+    kick: () => void;
+    /** Clear all pending timers once the request lifecycle is over. */
+    dispose: () => void;
+}
+/**
+ * Budget-aware stream abort controller. Combines an idle window (kicked on
+ * every received chunk — covers slow TTFT and hung connections) with a generous
+ * total hard cap (defaults to 420s / 7 minutes).
+ *
+ * Configurable via options or process.env:
+ *   - ARTIFACT_STREAM_IDLE_TIMEOUT_MS / STREAM_IDLE_TIMEOUT_MS
+ *   - ARTIFACT_STREAM_TOTAL_TIMEOUT_MS / STREAM_TOTAL_TIMEOUT_MS
+ */
+declare function createStreamAbortController(options?: {
+    idleMs?: number;
+    totalMs?: number;
+}): StreamAbortController;
+/** Backward-compatible alias */
+declare function createIdleAbortController(idleMs?: number, totalMs?: number): StreamAbortController;
+type IdleAbort = StreamAbortController;
 /**
  * Real-time Streaming Multi-Provider Inference Runner with Thinking & Tool Streaming
  */
@@ -76,4 +105,4 @@ declare function streamCurriculumAIInference(messages: ChatMessage[], projectCon
  */
 declare function runCurriculumAIInference(messages: ChatMessage[], projectContext: string, options?: StreamRunnerOptions, onChunk?: (token: string, type: ChunkType) => void): Promise<string>;
 
-export { type ChatMessage as C, type FallbackTarget as F, type StreamRunnerOptions as S, isModelAllowed as a, getDesignatedFallbackConfig as b, type ChunkType as c, extractThoughtAndContent as e, getDesignatedFallbackChain as g, isProviderEnabled as i, runCurriculumAIInference as r, streamCurriculumAIInference as s };
+export { type ChatMessage as C, DEFAULT_ARTIFACT_STREAM_IDLE_MS as D, type FallbackTarget as F, type IdleAbort as I, type StreamRunnerOptions as S, isModelAllowed as a, getDesignatedFallbackConfig as b, type ChunkType as c, DEFAULT_ARTIFACT_STREAM_TOTAL_MS as d, extractThoughtAndContent as e, type StreamAbortController as f, getDesignatedFallbackChain as g, createStreamAbortController as h, isProviderEnabled as i, createIdleAbortController as j, runCurriculumAIInference as r, streamCurriculumAIInference as s };

package/dist/{streamRunner-CcpmxOf1.d.ts → streamRunner-DtQq0HV_.d.ts}

@@ -15,6 +15,10 @@ interface StreamRunnerOptions {
     maxTokens?: number;
     temperature?: number;
     reasoningEffort?: 'off' | 'low' | 'medium' | 'high' | 'max';
+    /** Thời gian tối đa cho phép không nhận thêm token mới trước khi abort (mặc định 45_000ms = 45s) */
+    idleTimeoutMs?: number;
+    /** Thời gian hard cap tối đa cho toàn bộ lượt stream trước khi abort (mặc định 420_000ms = 7 phút) */
+    timeoutMs?: number;
     /**
      * Host-provided inference override (Vercel AI SDK gateway, test stubs, …).
      * When set, ALL built-in provider routes (raw-fetch NVIDIA/OpenRouter/
@@ -67,6 +71,31 @@ declare function extractThoughtAndContent(rawText: string): {
     thought: string;
     content: string;
 };
+declare const DEFAULT_ARTIFACT_STREAM_IDLE_MS = 45000;
+declare const DEFAULT_ARTIFACT_STREAM_TOTAL_MS = 420000;
+interface StreamAbortController {
+    signal: AbortSignal;
+    /** Reset the idle window (call on every received chunk). */
+    kick: () => void;
+    /** Clear all pending timers once the request lifecycle is over. */
+    dispose: () => void;
+}
+/**
+ * Budget-aware stream abort controller. Combines an idle window (kicked on
+ * every received chunk — covers slow TTFT and hung connections) with a generous
+ * total hard cap (defaults to 420s / 7 minutes).
+ *
+ * Configurable via options or process.env:
+ *   - ARTIFACT_STREAM_IDLE_TIMEOUT_MS / STREAM_IDLE_TIMEOUT_MS
+ *   - ARTIFACT_STREAM_TOTAL_TIMEOUT_MS / STREAM_TOTAL_TIMEOUT_MS
+ */
+declare function createStreamAbortController(options?: {
+    idleMs?: number;
+    totalMs?: number;
+}): StreamAbortController;
+/** Backward-compatible alias */
+declare function createIdleAbortController(idleMs?: number, totalMs?: number): StreamAbortController;
+type IdleAbort = StreamAbortController;
 /**
  * Real-time Streaming Multi-Provider Inference Runner with Thinking & Tool Streaming
  */
@@ -76,4 +105,4 @@ declare function streamCurriculumAIInference(messages: ChatMessage[], projectCon
  */
 declare function runCurriculumAIInference(messages: ChatMessage[], projectContext: string, options?: StreamRunnerOptions, onChunk?: (token: string, type: ChunkType) => void): Promise<string>;
 
-export { type ChatMessage as C, type FallbackTarget as F, type StreamRunnerOptions as S, isModelAllowed as a, getDesignatedFallbackConfig as b, type ChunkType as c, extractThoughtAndContent as e, getDesignatedFallbackChain as g, isProviderEnabled as i, runCurriculumAIInference as r, streamCurriculumAIInference as s };
+export { type ChatMessage as C, DEFAULT_ARTIFACT_STREAM_IDLE_MS as D, type FallbackTarget as F, type IdleAbort as I, type StreamRunnerOptions as S, isModelAllowed as a, getDesignatedFallbackConfig as b, type ChunkType as c, DEFAULT_ARTIFACT_STREAM_TOTAL_MS as d, extractThoughtAndContent as e, type StreamAbortController as f, getDesignatedFallbackChain as g, createStreamAbortController as h, isProviderEnabled as i, createIdleAbortController as j, runCurriculumAIInference as r, streamCurriculumAIInference as s };

package/dist/workflow/index.cjs

@@ -271,26 +271,38 @@ function extractThoughtAndContent(rawText) {
     content: content.trim()
   };
 }
-function createIdleAbortController(idleMs) {
+function createStreamAbortController(options) {
+  const envIdle = Number(process.env.ARTIFACT_STREAM_IDLE_TIMEOUT_MS || process.env.STREAM_IDLE_TIMEOUT_MS);
+  const envTotal = Number(process.env.ARTIFACT_STREAM_TOTAL_TIMEOUT_MS || process.env.STREAM_TOTAL_TIMEOUT_MS);
+  const idleMs = Number.isFinite(options?.idleMs) && options.idleMs > 0 ? options.idleMs : Number.isFinite(envIdle) && envIdle > 0 ? envIdle : DEFAULT_ARTIFACT_STREAM_IDLE_MS;
+  const totalMs = Number.isFinite(options?.totalMs) && options.totalMs > 0 ? options.totalMs : Number.isFinite(envTotal) && envTotal > 0 ? envTotal : DEFAULT_ARTIFACT_STREAM_TOTAL_MS;
   const controller = new AbortController();
-  let timer = null;
-  const arm = () => {
-    if (timer) clearTimeout(timer);
-    timer = setTimeout(
+  let idleTimer = null;
+  let totalTimer = null;
+  const armIdle = () => {
+    if (idleTimer) clearTimeout(idleTimer);
+    idleTimer = setTimeout(
       () => controller.abort(new Error(`Idle timeout: no data for ${Math.round(idleMs / 1e3)}s`)),
       idleMs
     );
-    timer?.unref?.();
+    idleTimer?.unref?.();
   };
-  arm();
+  armIdle();
+  if (totalMs > 0) {
+    totalTimer = setTimeout(
+      () => controller.abort(new Error(`Total stream timeout after ${Math.round(totalMs / 1e3)}s`)),
+      totalMs
+    );
+    totalTimer?.unref?.();
+  }
   return {
     signal: controller.signal,
-    /** Reset the idle window (call on every received chunk). */
-    kick: arm,
-    /** Clear the pending timer once the request lifecycle is over. */
+    kick: armIdle,
     dispose: () => {
-      if (timer) clearTimeout(timer);
-      timer = null;
+      if (idleTimer) clearTimeout(idleTimer);
+      if (totalTimer) clearTimeout(totalTimer);
+      idleTimer = null;
+      totalTimer = null;
     }
   };
 }
@@ -477,7 +489,7 @@ Guidelines:
     if (isNvidiaPreferred || uniqueNvidiaModels.length > 0) {
       for (const model of uniqueNvidiaModels) {
         try {
-          const idle = createIdleAbortController(18e4);
+          const idle = createStreamAbortController({ idleMs: options.idleTimeoutMs, totalMs: options.timeoutMs });
           const res = await fetch("https://integrate.api.nvidia.com/v1/chat/completions", {
             method: "POST",
             headers: {
@@ -521,7 +533,7 @@ Guidelines:
     const uniqueOrModels = Array.from(new Set(openrouterModels));
     for (const model of uniqueOrModels) {
       try {
-        const idle = createIdleAbortController(18e4);
+        const idle = createStreamAbortController({ idleMs: options.idleTimeoutMs, totalMs: options.timeoutMs });
         const res = await fetch("https://openrouter.ai/api/v1/chat/completions", {
           method: "POST",
           headers: {
@@ -560,7 +572,7 @@ Guidelines:
     const dsModel = options.model?.includes("deepseek-reasoner") ? "deepseek-reasoner" : "deepseek-chat";
     if (isModelAllowed(dsModel)) {
       try {
-        const idle = createIdleAbortController(18e4);
+        const idle = createStreamAbortController({ idleMs: options.idleTimeoutMs, totalMs: options.timeoutMs });
         const res = await fetch("https://api.deepseek.com/chat/completions", {
           method: "POST",
           headers: {
@@ -595,7 +607,7 @@ Guidelines:
     const baseUrl = alibabaKey.startsWith("sk-sp-") ? "https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1/chat/completions" : "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions";
     for (const qwenModel of qwenModels) {
       try {
-        const idle = createIdleAbortController(18e4);
+        const idle = createStreamAbortController({ idleMs: options.idleTimeoutMs, totalMs: options.timeoutMs });
         const payload = {
           model: qwenModel,
           messages: formattedMessages,
@@ -638,7 +650,7 @@ Guidelines:
     ];
     for (const gModel of geminiModels) {
       try {
-        const idle = createIdleAbortController(18e4);
+        const idle = createStreamAbortController({ idleMs: options.idleTimeoutMs, totalMs: options.timeoutMs });
         const res = await fetch(`https://generativelanguage.googleapis.com/v1beta/models/${gModel}:streamGenerateContent?alt=sse&key=${geminiKey}`, {
           method: "POST",
           headers: { "Content-Type": "application/json" },
@@ -674,7 +686,7 @@ Guidelines:
       const { provider, model } = target;
       if (provider === "nvidia" && nvidiaKey && isProviderEnabled("nvidia")) {
         try {
-          const idle = createIdleAbortController(18e4);
+          const idle = createStreamAbortController({ idleMs: options.idleTimeoutMs, totalMs: options.timeoutMs });
           const res = await fetch("https://integrate.api.nvidia.com/v1/chat/completions", {
             method: "POST",
             headers: {
@@ -706,7 +718,7 @@ Guidelines:
         }
       } else if (provider === "openrouter" && openrouterKey && isProviderEnabled("openrouter")) {
         try {
-          const idle = createIdleAbortController(18e4);
+          const idle = createStreamAbortController({ idleMs: options.idleTimeoutMs, totalMs: options.timeoutMs });
           const res = await fetch("https://openrouter.ai/api/v1/chat/completions", {
             method: "POST",
             headers: {
@@ -741,7 +753,7 @@ Guidelines:
       } else if ((provider === "alibaba" || provider === "dashscope") && alibabaKey && isProviderEnabled("alibaba")) {
         const baseUrl = alibabaKey.startsWith("sk-sp-") ? "https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1/chat/completions" : "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions";
         try {
-          const idle = createIdleAbortController(18e4);
+          const idle = createStreamAbortController({ idleMs: options.idleTimeoutMs, totalMs: options.timeoutMs });
           const res = await fetch(baseUrl, {
             method: "POST",
             headers: {
@@ -773,7 +785,7 @@ Guidelines:
         }
       } else if ((provider === "google" || provider === "gemini") && geminiKey && (isProviderEnabled("gemini") || isProviderEnabled("google"))) {
         try {
-          const idle = createIdleAbortController(18e4);
+          const idle = createStreamAbortController({ idleMs: options.idleTimeoutMs, totalMs: options.timeoutMs });
           const res = await fetch(
             `https://generativelanguage.googleapis.com/v1beta/models/${model}:streamGenerateContent?alt=sse&key=${geminiKey}`,
             {
@@ -821,9 +833,12 @@ Guidelines:
     rawError: providerErrors
   });
 }
+var DEFAULT_ARTIFACT_STREAM_IDLE_MS, DEFAULT_ARTIFACT_STREAM_TOTAL_MS;
 var init_streamRunner = __esm({
   "src/ai/streamRunner.ts"() {
     init_errors();
+    DEFAULT_ARTIFACT_STREAM_IDLE_MS = 45e3;
+    DEFAULT_ARTIFACT_STREAM_TOTAL_MS = 42e4;
   }
 });