@gobing-ai/knowledge-kit 0.0.16 → 0.0.18

package/plugins/generations/daily-article-gen/src/index.ts

@@ -8,7 +8,8 @@ import { echoError } from '@gobing-ai/ts-utils';
  * daily-article-gen: render a merged `Doc[]` into one readable daily AI-news
  * article. Mechanical template fill — no LLM, no STORM. Every news item keeps
  * a human-clickable source link (`[来源](sourceUri)`) so readers can trace
- * claims to their origin.
+ * claims to their origin. Research scaffolding is stripped before render
+ * (task 0138 R4b — see `stripScaffold`).
  *
  * Run date: `ARTICLE_DATE` env var (YYYYMMDD or ISO), else today (UTC).
  * Language-neutral: the Doc bodies already carry the summary language.
@@ -26,6 +27,71 @@ function sourceLabelOf(doc: Doc): string {
     }
 }
 
+/**
+ * Research-scaffold labels that must never reach reader copy (task 0138 R4b). The last30days
+ * research style emits them as `**「label」**` prefixes; the plan-translate prompt forbids them,
+ * but this deterministic pass catches what the agent misses. Grows with sources.
+ */
+const SCAFFOLD_LABELS = ['可做角度', '为什么现在值得注意', '社区讨论', '背景', '潜在影响', '影响'];
+
+/** Self-referential hedging markers — research voice, not news copy (task 0138 R4b). */
+const HEDGING_MARKERS = ['读者宜另行核实', '读者最好', '仍待验证', '尚待验证', '需另行核实'];
+
+/**
+ * Length of a leading scaffold label in any shape the research bodies use — `**「背景」**`,
+ * `「背景」`, `**背景**`, `### 背景`, `背景：` — or 0 when the line does not open with a known label.
+ * The capture is greedy and unbounded on purpose: it must reach the whole label (so `可做` cannot
+ * satisfy it where `可做角度` is the real one) and must not impose a length the label list outgrows.
+ * The `SCAFFOLD_LABELS` membership check, not the regex, decides.
+ */
+function scaffoldPrefixLength(text: string): number {
+    const match = text.match(
+        /^(?:#{1,6}\s*)?(?:\*\*)?\s*[「『]?\s*([^」』*：:\n]{2,})\s*[」』]?\s*(?:\*\*)?\s*[：:]?\s*/,
+    );
+    if (match === null || !SCAFFOLD_LABELS.includes((match[1] ?? '').trim())) return 0;
+    return match[0].length;
+}
+
+/** Drop every leading scaffold label, keeping the prose after it (`**「可做角度」** 可做角度：X` -> `X`). */
+function collapseScaffoldLabels(text: string): string {
+    let rest = text;
+    for (;;) {
+        const next = rest.slice(scaffoldPrefixLength(rest));
+        if (next === rest) return rest;
+        rest = next;
+    }
+}
+
+/** Drop the sentences carrying a hedging marker; a line that is only hedging collapses to ''. */
+function stripHedging(text: string): string {
+    if (!HEDGING_MARKERS.some((marker) => text.includes(marker))) return text;
+    return text
+        .split(/(?<=。)/)
+        .filter((sentence) => !HEDGING_MARKERS.some((marker) => sentence.includes(marker)))
+        .join('')
+        .trim();
+}
+
+/**
+ * Strip research scaffolding from a Doc body before it reaches reader copy (task 0138 R4b):
+ * standalone label lines are dropped, inline label prefixes collapse to their prose, and hedging
+ * sentences are removed. A clean body passes through byte-identical.
+ */
+export function stripScaffold(body: string): string {
+    const kept: string[] = [];
+    for (const line of body.split('\n')) {
+        const trimmed = line.trim();
+        if (trimmed === '') {
+            kept.push(line); // paragraph breaks survive untouched
+            continue;
+        }
+        const collapsed = stripHedging(collapseScaffoldLabels(trimmed));
+        if (collapsed === '') continue;
+        kept.push(collapsed === trimmed ? line : collapsed);
+    }
+    return kept.join('\n');
+}
+
 /** Render one news item section: heading, source link, and summary body. */
 function renderDoc(doc: Doc, index: number): string {
     const heading = doc.title?.trim() || `新闻 ${index + 1}`;
@@ -33,7 +99,7 @@ function renderDoc(doc: Doc, index: number): string {
     const meta = doc.metadata as Record<string, unknown> | undefined;
     const publishedAt = typeof meta?.publishedAt === 'string' ? meta.publishedAt.slice(0, 10) : '';
     const metaLine = [link, publishedAt].filter(Boolean).join(' · ');
-    return `## ${index + 1}. ${heading}\n\n${metaLine ? `${metaLine}\n\n` : ''}${(doc.body ?? '').trim()}`;
+    return `## ${index + 1}. ${heading}\n\n${metaLine ? `${metaLine}\n\n` : ''}${stripScaffold(doc.body ?? '').trim()}`;
 }
 
 /** Render the article markdown for a merged Doc[] (pure — exported for tests). */
@@ -54,11 +120,11 @@ function tagsFor(docs: Doc[]): string[] {
     return [AI_NEWS_TAG, ...existing.filter((t) => t !== AI_NEWS_TAG)];
 }
 
-/** Build the article Content from a merged Doc[]. */
-export function docsToContent(docs: Doc[], date: string): Content {
+/** Build the article Content from a merged Doc[]; `bodyOverride` replaces the mechanical render (agent-written prose), still scaffold-stripped as a safety net. */
+export function docsToContent(docs: Doc[], date: string, bodyOverride?: string): Content {
     return ContentSchema.parse({
         title: `每日 AI 资讯 — ${date}`,
-        body: renderDailyArticle(docs, date),
+        body: bodyOverride ? stripScaffold(bodyOverride).trim() : renderDailyArticle(docs, date),
         format: 'markdown',
         references: docs.filter((d) => d.sourceUri).map((d) => ({ url: d.sourceUri, title: d.title, cite: d.id })),
         metadata: { generator: 'kk:daily-article', date, doc_count: docs.length, tags: tagsFor(docs) },
@@ -69,7 +135,15 @@ export async function processGeneratorIO(inputPath: string, outputPath: string):
     const docs = DocListSchema.parse(await readJsonFile(inputPath));
     const raw = process.env.ARTICLE_DATE ?? new Date().toISOString().slice(0, 10);
     const date = raw.length === 8 ? `${raw.slice(0, 4)}-${raw.slice(4, 6)}-${raw.slice(6, 8)}` : raw;
-    const content = docsToContent(docs, date);
+    // Agent-written prose body (workflow `article-write` state) replaces the mechanical
+    // per-item render; missing/empty env ⇒ mechanical render as before.
+    let bodyOverride: string | undefined;
+    const bodyFile = process.env.ARTICLE_BODY_FILE;
+    if (bodyFile && bodyFile.length > 0) {
+        bodyOverride = await Bun.file(bodyFile).text();
+        if (bodyOverride.trim() === '') bodyOverride = undefined;
+    }
+    const content = docsToContent(docs, date, bodyOverride);
     const fs = createNodeFileSystem();
     const outDir = dirname(outputPath);
     if (outDir && outDir !== '.') await fs.ensureDir(outDir);

package/plugins/generations/dailynews-gen/dist/index.js

@@ -14642,72 +14642,6 @@ var init_zod = __esm(() => {
   init_external();
 });
 
-// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.39/node_modules/@gobing-ai/ts-utils/dist/errors.js
-var ErrorCode;
-var init_errors3 = __esm(() => {
-  ErrorCode = {
-    NotFound: "NOT_FOUND",
-    Validation: "VALIDATION",
-    Conflict: "CONFLICT",
-    Internal: "INTERNAL"
-  };
-});
-
-// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.39/node_modules/@gobing-ai/ts-utils/dist/api-response.js
-var API_ERROR_CODES, ERROR_CODE_TO_HTTP, CLIENT_SAFE_CODES;
-var init_api_response = __esm(() => {
-  init_errors3();
-  API_ERROR_CODES = {
-    SUCCESS: 0,
-    NOT_FOUND: 404,
-    VALIDATION_ERROR: 422,
-    BAD_REQUEST: 400,
-    UNAUTHORIZED: 401,
-    FORBIDDEN: 403,
-    CONFLICT: 409,
-    INTERNAL_ERROR: 500
-  };
-  ERROR_CODE_TO_HTTP = {
-    [ErrorCode.NotFound]: API_ERROR_CODES.NOT_FOUND,
-    [ErrorCode.Validation]: API_ERROR_CODES.VALIDATION_ERROR,
-    [ErrorCode.Conflict]: API_ERROR_CODES.CONFLICT,
-    [ErrorCode.Internal]: API_ERROR_CODES.INTERNAL_ERROR
-  };
-  CLIENT_SAFE_CODES = new Set([ErrorCode.NotFound, ErrorCode.Validation, ErrorCode.Conflict]);
-});
-
-// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.39/node_modules/@gobing-ai/ts-utils/dist/cursor.js
-var init_cursor = () => {};
-
-// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.39/node_modules/@gobing-ai/ts-utils/dist/event-severity.js
-var init_event_severity = () => {};
-
-// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.39/node_modules/@gobing-ai/ts-utils/dist/output.js
-function processStream(name) {
-  const proc = globalThis.process;
-  const stream = proc?.[name];
-  if (stream === undefined) {
-    throw new Error(`No ${name} target available: set one via setDefaultOutputTargets or pass an explicit target`);
-  }
-  return stream;
-}
-function writeLine(message, target) {
-  target.write(`${message}
-`);
-}
-function echoError(message, target = defaultStderrTarget ?? processStream("stderr")) {
-  writeLine(message, target);
-}
-var defaultStderrTarget;
-
-// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.39/node_modules/@gobing-ai/ts-utils/dist/index.js
-var init_dist = __esm(() => {
-  init_api_response();
-  init_cursor();
-  init_errors3();
-  init_event_severity();
-});
-
 // ../../node_modules/.bun/yaml@2.9.0/node_modules/yaml/dist/nodes/identity.js
 var require_identity = __commonJS((exports) => {
   var ALIAS = Symbol.for("yaml.alias");
@@ -17612,15 +17546,15 @@ var require_timestamp = __commonJS((exports) => {
         throw new Error("!!timestamp expects a date, starting with yyyy-mm-dd");
       const [, year, month, day, hour, minute, second] = match.map(Number);
       const millisec = match[7] ? Number((match[7] + "00").substr(1, 3)) : 0;
-      let date6 = Date.UTC(year, month - 1, day, hour || 0, minute || 0, second || 0, millisec);
+      let date5 = Date.UTC(year, month - 1, day, hour || 0, minute || 0, second || 0, millisec);
       const tz = match[8];
       if (tz && tz !== "Z") {
         let d = parseSexagesimal(tz, false);
         if (Math.abs(d) < 30)
           d *= 60;
-        date6 -= 60000 * d;
+        date5 -= 60000 * d;
       }
-      return new Date(date6);
+      return new Date(date5);
     },
     stringify: ({ value }) => value?.toISOString().replace(/(T00:00:00)?\.000Z$/, "") ?? ""
   };
@@ -19480,7 +19414,7 @@ var require_composer = __commonJS((exports) => {
   var node_process = __require("process");
   var directives = require_directives();
   var Document = require_Document();
-  var errors4 = require_errors();
+  var errors3 = require_errors();
   var identity = require_identity();
   var composeDoc = require_compose_doc();
   var resolveEnd = require_resolve_end();
@@ -19531,9 +19465,9 @@ var require_composer = __commonJS((exports) => {
       this.onError = (source, code, message, warning) => {
         const pos = getErrorPos(source);
         if (warning)
-          this.warnings.push(new errors4.YAMLWarning(pos, code, message));
+          this.warnings.push(new errors3.YAMLWarning(pos, code, message));
         else
-          this.errors.push(new errors4.YAMLParseError(pos, code, message));
+          this.errors.push(new errors3.YAMLParseError(pos, code, message));
       };
       this.directives = new directives.Directives({ version: options.version || "1.2" });
       this.options = options;
@@ -19619,7 +19553,7 @@ ${cb}` : comment;
           break;
         case "error": {
           const msg = token.source ? `${token.message}: ${JSON.stringify(token.source)}` : token.message;
-          const error51 = new errors4.YAMLParseError(getErrorPos(token), "UNEXPECTED_TOKEN", msg);
+          const error51 = new errors3.YAMLParseError(getErrorPos(token), "UNEXPECTED_TOKEN", msg);
           if (this.atDirectives || !this.doc)
             this.errors.push(error51);
           else
@@ -19629,7 +19563,7 @@ ${cb}` : comment;
         case "doc-end": {
           if (!this.doc) {
             const msg = "Unexpected doc-end without preceding document";
-            this.errors.push(new errors4.YAMLParseError(getErrorPos(token), "UNEXPECTED_TOKEN", msg));
+            this.errors.push(new errors3.YAMLParseError(getErrorPos(token), "UNEXPECTED_TOKEN", msg));
             break;
           }
           this.doc.directives.docEnd = true;
@@ -19644,7 +19578,7 @@ ${end.comment}` : end.comment;
           break;
         }
         default:
-          this.errors.push(new errors4.YAMLParseError(getErrorPos(token), "UNEXPECTED_TOKEN", `Unsupported token ${token.type}`));
+          this.errors.push(new errors3.YAMLParseError(getErrorPos(token), "UNEXPECTED_TOKEN", `Unsupported token ${token.type}`));
       }
     }
     *end(forceDoc = false, endOffset = -1) {
@@ -19670,7 +19604,7 @@ ${end.comment}` : end.comment;
 var require_cst_scalar = __commonJS((exports) => {
   var resolveBlockScalar = require_resolve_block_scalar();
   var resolveFlowScalar = require_resolve_flow_scalar();
-  var errors4 = require_errors();
+  var errors3 = require_errors();
   var stringifyString = require_stringifyString();
   function resolveAsScalar(token, strict = true, onError) {
     if (token) {
@@ -19679,7 +19613,7 @@ var require_cst_scalar = __commonJS((exports) => {
         if (onError)
           onError(offset, code, message);
         else
-          throw new errors4.YAMLParseError([offset, offset + 1], code, message);
+          throw new errors3.YAMLParseError([offset, offset + 1], code, message);
       };
       switch (token.type) {
         case "scalar":
@@ -21559,7 +21493,7 @@ var require_parser = __commonJS((exports) => {
 var require_public_api = __commonJS((exports) => {
   var composer = require_composer();
   var Document = require_Document();
-  var errors4 = require_errors();
+  var errors3 = require_errors();
   var log = require_log();
   var identity = require_identity();
   var lineCounter = require_line_counter();
@@ -21576,8 +21510,8 @@ var require_public_api = __commonJS((exports) => {
     const docs = Array.from(composer$1.compose(parser$1.parse(source)));
     if (prettyErrors && lineCounter2)
       for (const doc2 of docs) {
-        doc2.errors.forEach(errors4.prettifyError(source, lineCounter2));
-        doc2.warnings.forEach(errors4.prettifyError(source, lineCounter2));
+        doc2.errors.forEach(errors3.prettifyError(source, lineCounter2));
+        doc2.warnings.forEach(errors3.prettifyError(source, lineCounter2));
       }
     if (docs.length > 0)
       return docs;
@@ -21592,13 +21526,13 @@ var require_public_api = __commonJS((exports) => {
       if (!doc2)
         doc2 = _doc;
       else if (doc2.options.logLevel !== "silent") {
-        doc2.errors.push(new errors4.YAMLParseError(_doc.range.slice(0, 2), "MULTIPLE_DOCS", "Source contains multiple documents; please use YAML.parseAllDocuments()"));
+        doc2.errors.push(new errors3.YAMLParseError(_doc.range.slice(0, 2), "MULTIPLE_DOCS", "Source contains multiple documents; please use YAML.parseAllDocuments()"));
         break;
       }
     }
     if (prettyErrors && lineCounter2) {
-      doc2.errors.forEach(errors4.prettifyError(source, lineCounter2));
-      doc2.warnings.forEach(errors4.prettifyError(source, lineCounter2));
+      doc2.errors.forEach(errors3.prettifyError(source, lineCounter2));
+      doc2.warnings.forEach(errors3.prettifyError(source, lineCounter2));
     }
     return doc2;
   }
@@ -21650,12 +21584,12 @@ var require_public_api = __commonJS((exports) => {
 });
 
 // ../../node_modules/.bun/yaml@2.9.0/node_modules/yaml/dist/index.js
-var composer, Document, Schema, errors4, Alias, identity, Pair, Scalar, YAMLMap, YAMLSeq, cst, lexer, lineCounter, parser, publicApi, visit, $Composer, $Document, $Schema, $YAMLError, $YAMLParseError, $YAMLWarning, $Alias, $isAlias, $isCollection, $isDocument, $isMap, $isNode, $isPair, $isScalar, $isSeq, $Pair, $Scalar, $YAMLMap, $YAMLSeq, $Lexer, $LineCounter, $Parser, $parse, $parseAllDocuments, $parseDocument, $stringify, $visit, $visitAsync;
-var init_dist2 = __esm(() => {
+var composer, Document, Schema, errors3, Alias, identity, Pair, Scalar, YAMLMap, YAMLSeq, cst, lexer, lineCounter, parser, publicApi, visit, $Composer, $Document, $Schema, $YAMLError, $YAMLParseError, $YAMLWarning, $Alias, $isAlias, $isCollection, $isDocument, $isMap, $isNode, $isPair, $isScalar, $isSeq, $Pair, $Scalar, $YAMLMap, $YAMLSeq, $Lexer, $LineCounter, $Parser, $parse, $parseAllDocuments, $parseDocument, $stringify, $visit, $visitAsync;
+var init_dist = __esm(() => {
   composer = require_composer();
   Document = require_Document();
   Schema = require_Schema();
-  errors4 = require_errors();
+  errors3 = require_errors();
   Alias = require_Alias();
   identity = require_identity();
   Pair = require_Pair();
@@ -21671,9 +21605,9 @@ var init_dist2 = __esm(() => {
   $Composer = composer.Composer;
   $Document = Document.Document;
   $Schema = Schema.Schema;
-  $YAMLError = errors4.YAMLError;
-  $YAMLParseError = errors4.YAMLParseError;
-  $YAMLWarning = errors4.YAMLWarning;
+  $YAMLError = errors3.YAMLError;
+  $YAMLParseError = errors3.YAMLParseError;
+  $YAMLWarning = errors3.YAMLWarning;
   $Alias = Alias.Alias;
   $isAlias = identity.isAlias;
   $isCollection = identity.isCollection;
@@ -21701,7 +21635,7 @@ var init_dist2 = __esm(() => {
 // ../../node_modules/.bun/@gobing-ai+ts-runtime@0.4.39/node_modules/@gobing-ai/ts-runtime/dist/config.js
 var configSchema;
 var init_config = __esm(() => {
-  init_dist2();
+  init_dist();
   init_zod();
   configSchema = exports_external.object({
     app: exports_external.object({
@@ -21887,11 +21821,26 @@ var PluginManifestSchema = exports_external.object({
 init_file_system_node();
 init_config();
 
+// ../../node_modules/.bun/@gobing-ai+ts-utils@0.4.67/node_modules/@gobing-ai/ts-utils/dist/output.js
+var defaultStderrTarget;
+function processStream(name) {
+  const proc = globalThis.process;
+  const stream = proc?.[name];
+  if (stream === undefined) {
+    throw new Error(`No ${name} target available: set one via setDefaultOutputTargets or pass an explicit target`);
+  }
+  return stream;
+}
+function writeLine(message, target) {
+  target.write(`${message}
+`);
+}
+function echoError(message, target = defaultStderrTarget ?? processStream("stderr")) {
+  writeLine(message, target);
+}
+
 // ../../packages/kk-core/src/render-md.ts
 if (false) {}
-// ../../plugins/generations/dailynews-gen/src/index.ts
-init_dist();
-
 // ../../plugins/generations/dailynews-gen/src/script-builder.ts
 init_zod();
 var EffectConfigSchema = exports_external.object({
@@ -21922,6 +21871,7 @@ var VoiceScriptSegmentSchema = exports_external.object({
   engine: exports_external.string().min(1).optional(),
   language: exports_external.string().min(1).optional(),
   gap_ms: exports_external.number().int().nonnegative().optional(),
+  speed: exports_external.number().min(0.5).max(2).optional(),
   max_chunk_chars: exports_external.number().int().optional(),
   crossfade_ms: exports_external.number().int().optional(),
   personality: exports_external.boolean().optional(),
@@ -21951,13 +21901,13 @@ function formatChineseCount(n) {
     return ZH_DIGITS[n] ?? String(n);
   return String(n);
 }
-function formatBroadcastDate(date6, language) {
+function formatBroadcastDate(date5, language) {
   if (language === "zh") {
-    const year = date6.getFullYear();
-    const month = date6.getMonth() + 1;
-    const day = date6.getDate();
+    const year = date5.getFullYear();
+    const month = date5.getMonth() + 1;
+    const day = date5.getDate();
     const weekdays = ["\u65E5", "\u4E00", "\u4E8C", "\u4E09", "\u56DB", "\u4E94", "\u516D"];
-    const weekday = weekdays[date6.getDay()] ?? "\u65E5";
+    const weekday = weekdays[date5.getDay()] ?? "\u65E5";
     return `${year}\u5E74${month}\u6708${day}\u65E5\uFF0C\u661F\u671F${weekday}`;
   }
   return new Intl.DateTimeFormat("en-US", {
@@ -21965,7 +21915,7 @@ function formatBroadcastDate(date6, language) {
     month: "long",
     day: "numeric",
     year: "numeric"
-  }).format(date6);
+  }).format(date5);
 }
 function normalizeBroadcastText(raw, isZh) {
   let text = raw.trim();
@@ -21999,34 +21949,85 @@ function normalizeBroadcastText(raw, isZh) {
   }
   return text;
 }
-function synthesizePersonalUnderstanding(title, body, isZh) {
+function synthesizePersonalUnderstanding(title, body, isZh, used) {
   const combined = `${title} ${body}`.toLowerCase();
-  if (isZh) {
-    if (/jakub|szymon|\u9996\u5E2D\u79D1\u5B66\u5BB6|\u521B\u59CB\u56E2\u961F|\u6280\u672F\u9886\u8896|\u9EC4\u91D1\u642D\u6863/i.test(combined)) {
-      return "\u5728\u5404\u79CD\u6280\u672F\u5149\u73AF\u80CC\u540E\uFF0C\u6B63\u662F\u8FD9\u6837\u4E13\u6CE8\u800C\u4E0D\u77E5\u75B2\u5026\u7684\u5E95\u5C42\u653B\u575A\u8005\uFF0C\u4E3A\u4ECA\u5929\u6574\u4E2A\u5927\u6A21\u578B\u751F\u6001\u5960\u5B9A\u4E86\u6700\u575A\u5B9E\u7684\u6280\u672F\u57FA\u77F3\u3002";
-    }
-    if (/\u7B97\u529B|\u5409\u74E6|\u5DE5\u5382|\u7535\u529B|\u82AF\u7247|\u57FA\u7840\u8BBE\u65BD|\u57FA\u5EFA|\u786C\u4EF6|\u80FD\u6E90/i.test(combined)) {
-      return "\u8FD9\u5176\u5B9E\u91CA\u653E\u4E86\u4E00\u4E2A\u975E\u5E38\u660E\u786E\u7684\u4FE1\u53F7\uFF1AAI\u7ADE\u4E89\u7684\u4E0B\u534A\u573A\uFF0C\u5DF2\u7ECF\u4E0D\u518D\u53EA\u662F\u5355\u7EAF\u62FC\u6A21\u578B\u53C2\u6570\uFF0C\u66F4\u662F\u7535\u529B\u3001\u80FD\u6E90\u548C\u5168\u6808\u5DE5\u4E1A\u57FA\u7840\u8BBE\u65BD\u7684\u786C\u6838\u8F83\u91CF\u3002";
-    }
-    if (/sora|\u89C6\u9891|\u751F\u6210|cameo|chatgpt/i.test(combined)) {
-      return "\u6211\u4E2A\u4EBA\u89C9\u5F97\uFF0C\u8FD9\u6B21\u6700\u5927\u7684\u770B\u70B9\u662F\u5185\u5BB9\u521B\u4F5C\u95E8\u69DB\u7684\u518D\u6B21\u964D\u4F4E\u3002\u4E0D\u8FC7\u6B63\u5982\u56E2\u961F\u63D0\u5230\u7684\uFF0C\u5982\u4F55\u5E73\u8861\u521B\u4F5C\u81EA\u7531\u4E0E\u9632\u6B62\u4F4E\u8D28\u5185\u5BB9\u6CDB\u6EE5\uFF0C\u4F1A\u662F\u6574\u4E2A\u884C\u4E1A\u63A5\u4E0B\u6765\u6700\u503C\u5F97\u89C2\u5BDF\u7684\u5173\u952E\u8003\u9898\u3002";
-    }
-    if (/\u878D\u8D44|\u6295\u8D44|\u5E02\u573A|\u5546\u4E1A|\u4F30\u503C|\u5408\u4F5C/i.test(combined)) {
-      return "\u8FD9\u9879\u8FDB\u5C55\u53CD\u6620\u51FA\u8D44\u672C\u4E0E\u4EA7\u4E1A\u754C\u6B63\u5728\u52A0\u901F\u5F62\u6210\u5408\u529B\uFF0C\u540E\u7EED\u5B9E\u9645\u4E1A\u52A1\u573A\u666F\u7684\u843D\u5730\u6548\u679C\u975E\u5E38\u503C\u5F97\u6211\u4EEC\u6301\u7EED\u8DDF\u8FDB\u3002";
-    }
-    return "\u8FD9\u9879\u65B0\u52A8\u6001\u53CD\u6620\u51FA\u6280\u672F\u6F14\u8FDB\u6B63\u5728\u8FDB\u4E00\u6B65\u63D0\u901F\uFF0C\u5BF9\u76F8\u5173\u9886\u57DF\u7684\u5B9E\u9645\u5F71\u54CD\u4E5F\u975E\u5E38\u503C\u5F97\u6211\u4EEC\u6301\u7EED\u5173\u6CE8\u3002";
+  const line = isZh ? zhInsightFor(combined) : enInsightFor(combined);
+  if (used.has(line))
+    return null;
+  used.add(line);
+  return line;
+}
+function zhInsightFor(combined) {
+  if (/jakub|szymon|\u9996\u5E2D\u79D1\u5B66\u5BB6|\u521B\u59CB\u56E2\u961F|\u6280\u672F\u9886\u8896|\u9EC4\u91D1\u642D\u6863/i.test(combined)) {
+    return "\u5728\u5404\u79CD\u6280\u672F\u5149\u73AF\u80CC\u540E\uFF0C\u6B63\u662F\u8FD9\u6837\u4E13\u6CE8\u800C\u4E0D\u77E5\u75B2\u5026\u7684\u5E95\u5C42\u653B\u575A\u8005\uFF0C\u4E3A\u4ECA\u5929\u6574\u4E2A\u5927\u6A21\u578B\u751F\u6001\u5960\u5B9A\u4E86\u6700\u575A\u5B9E\u7684\u6280\u672F\u57FA\u77F3\u3002";
   }
+  if (/\u7B97\u529B|\u5409\u74E6|\u7535\u529B|\u80FD\u6E90|\u6570\u636E\u4E2D\u5FC3|\u57FA\u7840\u8BBE\u65BD|\u57FA\u5EFA/.test(combined)) {
+    return "\u8FD9\u5176\u5B9E\u91CA\u653E\u4E86\u4E00\u4E2A\u975E\u5E38\u660E\u786E\u7684\u4FE1\u53F7\uFF1AAI\u7ADE\u4E89\u7684\u4E0B\u534A\u573A\uFF0C\u5DF2\u7ECF\u4E0D\u518D\u53EA\u662F\u5355\u7EAF\u62FC\u6A21\u578B\u53C2\u6570\uFF0C\u66F4\u662F\u7535\u529B\u3001\u80FD\u6E90\u548C\u5168\u6808\u5DE5\u4E1A\u57FA\u7840\u8BBE\u65BD\u7684\u786C\u6838\u8F83\u91CF\u3002";
+  }
+  if (/\u82AF\u7247|\u786C\u4EF6|\u5904\u7406\u5668|enclave/i.test(combined)) {
+    return "\u771F\u6B63\u88AB\u91CD\u65B0\u5B9A\u4E49\u7684\uFF0C\u662F\u786C\u4EF6\u5C42\u627F\u62C5\u7684\u5B89\u5168\u4E0E\u7B97\u529B\u8D23\u4EFB\uFF1A\u628A\u80FD\u529B\u505A\u8FDB\u7845\u7247\u5C42\u66F4\u53EF\u9760\u3001\u4E5F\u66F4\u96BE\u7ED5\u8FC7\uFF0C\u4F46\u751F\u6001\u9501\u5B9A\u7684\u4EE3\u4EF7\u9700\u8981\u4E00\u8D77\u7B97\u8FDB\u53BB\u3002";
+  }
+  if (/sora|\u89C6\u9891|\u751F\u6210\u89C6\u9891|\u89C6\u9891\u751F\u6210|\u6587\u751F\u89C6\u9891|\u5185\u5BB9\u521B\u4F5C|cameo/i.test(combined)) {
+    return "\u6211\u4E2A\u4EBA\u89C9\u5F97\uFF0C\u8FD9\u6B21\u6700\u5927\u7684\u770B\u70B9\u662F\u5185\u5BB9\u521B\u4F5C\u95E8\u69DB\u7684\u518D\u6B21\u964D\u4F4E\u3002\u4E0D\u8FC7\u6B63\u5982\u56E2\u961F\u63D0\u5230\u7684\uFF0C\u5982\u4F55\u5E73\u8861\u521B\u4F5C\u81EA\u7531\u4E0E\u9632\u6B62\u4F4E\u8D28\u5185\u5BB9\u6CDB\u6EE5\uFF0C\u4F1A\u662F\u6574\u4E2A\u884C\u4E1A\u63A5\u4E0B\u6765\u6700\u503C\u5F97\u89C2\u5BDF\u7684\u5173\u952E\u8003\u9898\u3002";
+  }
+  if (/\u878D\u8D44|\u6295\u8D44|\u5E02\u573A|\u5546\u4E1A|\u4F30\u503C|\u5408\u4F5C/.test(combined)) {
+    return "\u8FD9\u9879\u8FDB\u5C55\u53CD\u6620\u51FA\u8D44\u672C\u4E0E\u4EA7\u4E1A\u754C\u6B63\u5728\u52A0\u901F\u5F62\u6210\u5408\u529B\uFF0C\u540E\u7EED\u5B9E\u9645\u4E1A\u52A1\u573A\u666F\u7684\u843D\u5730\u6548\u679C\u975E\u5E38\u503C\u5F97\u6211\u4EEC\u6301\u7EED\u8DDF\u8FDB\u3002";
+  }
+  return "\u8FD9\u9879\u65B0\u52A8\u6001\u53CD\u6620\u51FA\u6280\u672F\u6F14\u8FDB\u6B63\u5728\u8FDB\u4E00\u6B65\u63D0\u901F\uFF0C\u5BF9\u76F8\u5173\u9886\u57DF\u7684\u5B9E\u9645\u5F71\u54CD\u4E5F\u975E\u5E38\u503C\u5F97\u6211\u4EEC\u6301\u7EED\u5173\u6CE8\u3002";
+}
+function enInsightFor(combined) {
   if (/jakub|szymon|chief scientist|founding team|technical leader|partnership/i.test(combined)) {
     return "Behind the headlines, it is relentless technical leadership and deep engineering rigor that continue to drive these foundational milestones.";
   }
-  if (/compute|gigawatt|factory|power|chip|infrastructure|datacenter/i.test(combined)) {
+  if (/compute|gigawatt|power|energy|datacenter|infrastructure/.test(combined)) {
     return "This signals clearly that the AI frontier is no longer just about algorithms, but a massive contest of power, energy, and full-stack industrial engineering.";
   }
-  if (/sora|video|generation|cameo|chatgpt/i.test(combined)) {
+  if (/chip|hardware|processor|enclave/i.test(combined)) {
+    return "What stands out is the hardware layer quietly taking on the security and compute burden: putting the guarantee in silicon is harder to bypass, and it carries its own lock-in cost.";
+  }
+  if (/sora|video|video generation|content creation|cameo/i.test(combined)) {
     return "In my view, the biggest takeaway here is the dramatic lowering of creation barriers, while content quality governance will be the key test ahead.";
   }
   return "This development highlights how rapidly the ecosystem is moving from experimental proofs of concept into large-scale execution.";
 }
+var MAX_FACT_SEGMENT_CHARS = 300;
+var TITLE_ECHO_NOISE_RE = /[\uFF0C\u3002\u3001\uFF1A:\uFF1B;!\uFF01?\uFF1F\u2026\s\u300C\u300D\u300E\u300F\u3010\u3011]/g;
+function stripLeadingTitleEcho(body, title) {
+  const wanted = title.replace(TITLE_ECHO_NOISE_RE, "");
+  if (wanted.length < 4)
+    return body;
+  let matched = "";
+  let end = -1;
+  for (let i = 0;i < body.length && matched.length < wanted.length; i++) {
+    const kept = (body[i] ?? "").replace(TITLE_ECHO_NOISE_RE, "");
+    if (kept === "")
+      continue;
+    matched += kept;
+    if (matched.length >= wanted.length)
+      end = i + 1;
+  }
+  if (end === -1 || matched !== wanted)
+    return body;
+  return body.slice(end).replace(/^[\uFF0C\u3002\u3001\uFF1A:\uFF1B;!\uFF01?\uFF1F\u2026\s\u300C\u300D\u300E\u300F\u3010\u3011]+/, "");
+}
+function splitFactText(text, maxChars = MAX_FACT_SEGMENT_CHARS) {
+  if (text.length <= maxChars)
+    return [text];
+  const clauses = text.match(/[^\u3002\uFF01\uFF1F\u2026\uFF1B]+[\u3002\uFF01\uFF1F\u2026\uFF1B]*|[\s\S]/g) ?? [text];
+  const chunks = [];
+  let current = "";
+  for (const clause of clauses) {
+    if (current !== "" && current.length + clause.length > maxChars) {
+      chunks.push(current);
+      current = clause;
+      continue;
+    }
+    current += clause;
+  }
+  if (current !== "")
+    chunks.push(current);
+  return chunks.length > 0 ? chunks : [text];
+}
 function buildNewsVoiceScript(docs, options) {
   const language = exports_external.enum(["zh", "en"]).parse((options?.language || "zh").toLowerCase());
   const isZh = language === "zh";
@@ -22086,26 +22087,33 @@ function buildNewsVoiceScript(docs, options) {
     gap_ms: 700,
     emotion: "confident, warm"
   });
+  const usedInsights = new Set;
   docs.forEach((doc2, idx) => {
     const isLast = idx === docs.length - 1 && docs.length > 1;
     const cleanTitle = (doc2.title ?? "").trim();
-    const cleanBody = (doc2.body ?? "").trim();
+    const cleanBody = stripLeadingTitleEcho((doc2.body ?? "").trim(), cleanTitle);
     const lead = isZh ? idx === 0 ? `\u9996\u5148\u6765\u804A\u804A\u5927\u5BB6\u975E\u5E38\u5173\u6CE8\u7684\u3010${cleanTitle}\u3011\u3002` : isLast ? `\u6700\u540E\uFF0C\u6765\u770B\u770B\u4ECA\u5929\u7684\u6700\u540E\u4E00\u6761\u8D44\u8BAF\uFF0C\u3010${cleanTitle}\u3011\u3002` : `\u63A5\u7740\u6211\u4EEC\u628A\u76EE\u5149\u8F6C\u5411\u53E6\u4E00\u6761\u91CD\u8981\u8FDB\u5C55\uFF0C\u3010${cleanTitle}\u3011\u3002` : idx === 0 ? `First up today, let's look at ${cleanTitle}.` : isLast ? `And finally today, let's wrap up with ${cleanTitle}.` : `Next, turning our attention to ${cleanTitle}.`;
     const normalizedLead = normalizeBroadcastText(lead, isZh);
-    const normalizedBody = normalizeBroadcastText(cleanBody, isZh);
-    const insight = synthesizePersonalUnderstanding(cleanTitle, cleanBody, isZh);
-    const normalizedInsight = normalizeBroadcastText(insight, isZh);
+    const normalizedBody = cleanBody === "" ? "" : normalizeBroadcastText(cleanBody, isZh);
+    const insight = synthesizePersonalUnderstanding(cleanTitle, cleanBody, isZh, usedInsights);
+    const normalizedInsight = insight === null ? null : normalizeBroadcastText(insight, isZh);
     const factText = isZh ? `${normalizedLead}${normalizedBody}` : `${normalizedLead} ${normalizedBody}`;
-    segments.push({
-      text: factText,
-      gap_ms: 500,
-      emotion: "professional, clear"
-    });
-    segments.push({
-      text: normalizedInsight,
-      gap_ms: 800,
-      emotion: "thoughtful, personal"
-    });
+    for (const chunk of splitFactText(factText)) {
+      segments.push({
+        text: chunk,
+        gap_ms: 500,
+        emotion: "professional, clear",
+        speed: 1
+      });
+    }
+    if (normalizedInsight !== null) {
+      segments.push({
+        text: normalizedInsight,
+        gap_ms: 1200,
+        emotion: "thoughtful, personal",
+        speed: 1.12
+      });
+    }
   });
   const outroText = isZh ? `\u597D\u4E86\uFF0C\u4EE5\u4E0A\u5C31\u662F\u4ECA\u5929AI\u79D1\u6280\u6F2B\u8C08\u7684\u5168\u90E8\u5185\u5BB9\u3002\u4E0D\u77E5\u9053\u4ECA\u5929\u8FD9\u51E0\u6761\u6D88\u606F\u6709\u6CA1\u6709\u7ED9\u60A8\u5E26\u6765\u4E00\u4E9B\u65B0\u7684\u542F\u53D1\uFF1F\u611F\u8C22\u60A8\u7684\u6536\u542C\uFF0C\u6211\u662F${hostName}\uFF0C\u795D\u60A8\u62E5\u6709\u5145\u5B9E\u9AD8\u6548\u7684\u4E00\u5929\uFF0C\u6211\u4EEC\u4E0B\u671F\u8282\u76EE\u4E0D\u89C1\u4E0D\u6563\uFF01` : `That wraps up today's AI Daily Briefing. I hope today's highlights gave you some useful takeaways. Thank you for listening. This is ${hostName}, wishing you a productive and wonderful day. Until next time!`;
   segments.push({
@@ -22216,6 +22224,9 @@ function formatVoiceScriptToYaml(script) {
     if (segment.gap_ms !== undefined) {
       lines.push(`    gap_ms: ${segment.gap_ms}`);
     }
+    if (segment.speed !== undefined) {
+      lines.push(`    speed: ${segment.speed}`);
+    }
     if (segment.seed !== undefined) {
       lines.push(`    seed: ${segment.seed}`);
     }
@@ -22330,6 +22341,8 @@ if (import.meta.main) {
 }
 export {
   synthesizePersonalUnderstanding,
+  stripLeadingTitleEcho,
+  splitFactText,
   processGeneratorIO,
   normalizeBroadcastText,
   main2 as main,
@@ -22340,5 +22353,6 @@ export {
   VoiceSpeakerSchema,
   VoiceScriptSegmentSchema,
   VoiceScriptSchema,
+  MAX_FACT_SEGMENT_CHARS,
   EffectConfigSchema
 };