publishport-opencli 1.8.5-pp.2 → 1.8.5-pp.21

package/clis/youtube/subscriptions.js

@@ -1,31 +1,9 @@
-/**
- * YouTube subscriptions — list of subscribed channels from /feed/channels.
- */
-import { cli, Strategy } from '@jackwener/opencli/registry';
-import { CommandExecutionError, EmptyResultError } from '@jackwener/opencli/errors';
-import { extractSubscriptionChannel } from './utils.js';
-
-cli({
-    site: 'youtube',
-    name: 'subscriptions',
-    access: 'read',
-    description: 'List subscribed YouTube channels',
-    domain: 'www.youtube.com',
-    strategy: Strategy.COOKIE,
-    args: [
-        { name: 'limit', type: 'int', default: 50, help: 'Max channels to return (default 50)' },
-    ],
-    columns: ['rank', 'name', 'handle', 'subscribers', 'url'],
-    func: async (page, kwargs) => {
-        const limit = Math.min(kwargs.limit || 50, 1000);
-        await page.goto('https://www.youtube.com/feed/channels');
-        await page.wait(3);
-        const data = await page.evaluate(`
+import{cli as i,Strategy as a}from"@jackwener/opencli/registry";import{CommandExecutionError as c,EmptyResultError as u}from"@jackwener/opencli/errors";import{extractSubscriptionChannel as l}from"./utils.js";i({site:"youtube",name:"subscriptions",access:"read",description:"List subscribed YouTube channels",domain:"www.youtube.com",strategy:a.COOKIE,args:[{name:"limit",type:"int",default:50,help:"Max channels to return (default 50)"}],columns:["rank","name","handle","subscribers","url"],func:async(t,r)=>{const o=Math.min(r.limit||50,1000);await t.goto("https://www.youtube.com/feed/channels");await t.wait(3);const e=await t.evaluate(`
       (async () => {
         const d = window.ytInitialData;
         if (!d) return { error: 'YouTube data not found — are you logged in?' };
 
-        const limit = ${limit};
+        const limit = ${o};
 
         const items = d.contents?.twoColumnBrowseResultsRenderer
           ?.tabs?.[0]?.tabRenderer?.content
@@ -34,7 +12,7 @@ cli({
           ?.shelfRenderer?.content
           ?.expandedShelfContentsRenderer?.items || [];
 
-        const extractChannel = ${extractSubscriptionChannel.toString()};
+        const extractChannel = ${l.toString()};
 
         const channels = [];
         for (const item of items) {
@@ -45,14 +23,4 @@ cli({
 
         return channels;
       })()
-    `);
-        if (!Array.isArray(data)) {
-            const errMsg = data && typeof data === 'object' ? String(data.error || '') : '';
-            throw new CommandExecutionError(errMsg || 'Failed to fetch subscriptions — make sure you are logged into YouTube');
-        }
-        if (data.length === 0) {
-            throw new EmptyResultError('youtube subscriptions');
-        }
-        return data.map((ch, i) => ({ rank: i + 1, ...ch }));
-    },
-});
+    `);if(!Array.isArray(e)){const n=e&&typeof e==="object"?String(e.error||""):"";throw new c(n||"Failed to fetch subscriptions — make sure you are logged into YouTube")}if(e.length===0)throw new u("youtube subscriptions");return e.map((n,s)=>({rank:s+1,...n}))}});

package/clis/youtube/transcript-group.js

@@ -1,226 +1,2 @@
-/**
- * Transcript grouping: sentence merging, speaker detection, and chapter support.
- * Ported and simplified from Defuddle's YouTube extractor.
- *
- * Raw segments (2-3 second fragments) are grouped into readable paragraphs:
- * - Sentence boundaries: merge until sentence-ending punctuation (.!?)
- * - Speaker turns: detect ">>" markers from YouTube auto-captions
- * - Chapters: optional chapter headings inserted at appropriate timestamps
- */
-// Include CJK sentence-ending punctuation: 。！？ (fullwidth: ．！？)
-const SENTENCE_END = /[.!?\u3002\uFF01\uFF1F\uFF0E]["'\u2019\u201D)]*\s*$/;
-const QUESTION_END = /[?\uFF1F]["'\u2019\u201D)]*\s*$/;
-const TRANSCRIPT_GROUP_GAP_SECONDS = 20;
-const TURN_MERGE_MAX_WORDS = 80;
-const TURN_MERGE_MAX_SPAN_SECONDS = 45;
-const SHORT_UTTERANCE_MAX_WORDS = 3;
-const FIRST_GROUP_MERGE_MIN_WORDS = 8;
-function countWords(text) {
-    return text.split(/\s+/).filter(Boolean).length;
-}
-/**
- * Group raw transcript segments into readable blocks.
- * If speaker markers (>>) are present, groups by speaker turn.
- * Otherwise, groups by sentence boundaries.
- */
-export function groupTranscriptSegments(segments) {
-    if (segments.length === 0)
-        return [];
-    const hasSpeakerMarkers = segments.some(s => /^>>/.test(s.text));
-    return hasSpeakerMarkers ? groupBySpeaker(segments) : groupBySentence(segments);
-}
-/**
- * Format grouped segments + chapters into a final text output.
- */
-export function formatGroupedTranscript(segments, chapters = []) {
-    const sortedChapters = [...chapters].sort((a, b) => a.start - b.start);
-    let chapterIdx = 0;
-    const rows = [];
-    const textParts = [];
-    for (const segment of segments) {
-        // Insert chapter headings
-        while (chapterIdx < sortedChapters.length && sortedChapters[chapterIdx].start <= segment.start) {
-            const title = sortedChapters[chapterIdx].title;
-            rows.push({ timestamp: fmtTime(sortedChapters[chapterIdx].start), speaker: '', text: `[Chapter] ${title}` });
-            if (textParts.length > 0)
-                textParts.push('');
-            textParts.push(`### ${title}`);
-            textParts.push('');
-            chapterIdx++;
-        }
-        const timestamp = fmtTime(segment.start);
-        const speaker = segment.speaker !== undefined ? `Speaker ${segment.speaker + 1}` : '';
-        rows.push({ timestamp, speaker, text: segment.text });
-        if (segment.speakerChange && textParts.length > 0) {
-            textParts.push('');
-        }
-        textParts.push(`${timestamp} ${segment.text}`);
-    }
-    return { rows, plainText: textParts.join('\n') };
-}
-function fmtTime(sec) {
-    const h = Math.floor(sec / 3600);
-    const m = Math.floor((sec % 3600) / 60);
-    const s = Math.floor(sec % 60);
-    if (h > 0) {
-        return `${h}:${String(m).padStart(2, '0')}:${String(s).padStart(2, '0')}`;
-    }
-    return `${m}:${String(s).padStart(2, '0')}`;
-}
-// ── Sentence grouping ─────────────────────────────────────────────────────
-// Max time span (seconds) for a single group when no sentence boundaries are found.
-// Prevents unbounded merging for languages without punctuation (Chinese, etc.).
-const MAX_GROUP_SPAN_SECONDS = 30;
-function groupBySentence(segments) {
-    const groups = [];
-    let buffer = '';
-    let bufferStart = 0;
-    let lastStart = 0;
-    const flush = () => {
-        if (buffer.trim()) {
-            groups.push({ start: bufferStart, text: buffer.trim(), speakerChange: false });
-            buffer = '';
-        }
-    };
-    for (const seg of segments) {
-        // Large gap between segments — always flush
-        if (buffer && seg.start - lastStart > TRANSCRIPT_GROUP_GAP_SECONDS) {
-            flush();
-        }
-        // Time-based flush: prevent unbounded groups for unpunctuated languages
-        if (buffer && seg.start - bufferStart > MAX_GROUP_SPAN_SECONDS) {
-            flush();
-        }
-        if (!buffer)
-            bufferStart = seg.start;
-        buffer += (buffer ? ' ' : '') + seg.text;
-        lastStart = seg.start;
-        if (SENTENCE_END.test(seg.text))
-            flush();
-    }
-    flush();
-    return groups;
-}
-// ── Speaker grouping ──────────────────────────────────────────────────────
-function groupBySpeaker(segments) {
-    const turns = [];
-    let currentTurn = null;
-    let speakerIndex = -1;
-    let prevSegText = '';
-    for (const seg of segments) {
-        const isSpeakerChange = /^>>/.test(seg.text);
-        const cleanText = seg.text.replace(/^>>\s*/, '').replace(/^-\s+/, '');
-        const prevEndsWithComma = /,\s*$/.test(prevSegText);
-        const prevEndedSentence = (SENTENCE_END.test(prevSegText) || !prevSegText) && !prevEndsWithComma;
-        const isRealSpeakerChange = isSpeakerChange && prevEndedSentence;
-        if (isRealSpeakerChange) {
-            if (currentTurn)
-                turns.push(currentTurn);
-            speakerIndex = (speakerIndex + 1) % 2;
-            currentTurn = {
-                start: seg.start,
-                segments: [{ start: seg.start, text: cleanText }],
-                speakerChange: true,
-                speaker: speakerIndex,
-            };
-        }
-        else {
-            if (!currentTurn) {
-                currentTurn = { start: seg.start, segments: [], speakerChange: false };
-            }
-            currentTurn.segments.push({ start: seg.start, text: cleanText });
-        }
-        prevSegText = cleanText;
-    }
-    if (currentTurn)
-        turns.push(currentTurn);
-    splitAffirmativeTurns(turns);
-    const groups = [];
-    for (const turn of turns) {
-        const sentenceGroups = turn.speaker === undefined
-            ? groupBySentence(turn.segments)
-            : mergeSentenceGroupsWithinTurn(groupBySentence(turn.segments));
-        for (let i = 0; i < sentenceGroups.length; i++) {
-            groups.push({
-                ...sentenceGroups[i],
-                speakerChange: i === 0 && turn.speakerChange,
-                speaker: turn.speaker,
-            });
-        }
-    }
-    return groups;
-}
-function splitAffirmativeTurns(turns) {
-    const affirmativePattern = /^(mhm|yeah|yes|yep|right|okay|ok|absolutely|sure|exactly|uh-huh|mm-hmm)[.!,]?\s+/i;
-    for (let i = 0; i < turns.length; i++) {
-        const turn = turns[i];
-        if (turn.speaker === undefined || turn.segments.length === 0)
-            continue;
-        const firstSeg = turn.segments[0];
-        const match = affirmativePattern.exec(firstSeg.text);
-        if (!match)
-            continue;
-        if (/,\s*$/.test(match[0]))
-            continue;
-        const remainder = firstSeg.text.slice(match[0].length).trim();
-        const restSegments = turn.segments.slice(1);
-        const restWords = countWords(remainder) + restSegments.reduce((sum, s) => sum + countWords(s.text), 0);
-        if (restWords < 30)
-            continue;
-        const affirmativeText = match[0].trimEnd();
-        const newRestSegments = remainder
-            ? [{ start: firstSeg.start, text: remainder }, ...restSegments]
-            : restSegments;
-        turns.splice(i, 1, {
-            start: turn.start,
-            segments: [{ start: firstSeg.start, text: affirmativeText }],
-            speakerChange: turn.speakerChange,
-            speaker: turn.speaker,
-        }, {
-            start: newRestSegments[0].start,
-            segments: newRestSegments,
-            speakerChange: true,
-            speaker: turn.speaker === 0 ? 1 : 0,
-        });
-        i++;
-    }
-}
-function mergeSentenceGroupsWithinTurn(groups) {
-    if (groups.length <= 1)
-        return groups;
-    const merged = [];
-    let current = { ...groups[0] };
-    let currentIsFirstInTurn = true;
-    for (let i = 1; i < groups.length; i++) {
-        const next = groups[i];
-        if (shouldMergeSentenceGroups(current, next, currentIsFirstInTurn)) {
-            current.text = `${current.text} ${next.text}`;
-            continue;
-        }
-        merged.push(current);
-        current = { ...next };
-        currentIsFirstInTurn = false;
-    }
-    merged.push(current);
-    return merged;
-}
-function shouldMergeSentenceGroups(current, next, currentIsFirstInTurn) {
-    const currentWords = countWords(current.text);
-    const nextWords = countWords(next.text);
-    if (isShortStandaloneUtterance(current.text, currentWords)
-        || isShortStandaloneUtterance(next.text, nextWords))
-        return false;
-    if (currentIsFirstInTurn && currentWords < FIRST_GROUP_MERGE_MIN_WORDS)
-        return false;
-    if (QUESTION_END.test(current.text) || QUESTION_END.test(next.text))
-        return false;
-    if (currentWords + nextWords > TURN_MERGE_MAX_WORDS)
-        return false;
-    if (next.start - current.start > TURN_MERGE_MAX_SPAN_SECONDS)
-        return false;
-    return true;
-}
-function isShortStandaloneUtterance(text, words) {
-    const w = words ?? countWords(text);
-    return w > 0 && w <= SHORT_UTTERANCE_MAX_WORDS && SENTENCE_END.test(text);
-}
+const y=/[.!?\u3002\uFF01\uFF1F\uFF0E]["'\u2019\u201D)]*\s*$/,F=/[?\uFF1F]["'\u2019\u201D)]*\s*$/,M=20,A=80,U=45,D=3,G=8;function H(q){return q.split(/\s+/).filter(Boolean).length}export function groupTranscriptSegments(q){if(q.length===0)return[];return q.some((j)=>/^>>/.test(j.text))?b(q):B(q)}export function formatGroupedTranscript(q,K=[]){const j=[...K].sort((z,Y)=>z.start-Y.start);let J=0;const V=[],L=[];for(const z of q){while(J<j.length&&j[J].start<=z.start){const $=j[J].title;V.push({timestamp:k(j[J].start),speaker:"",text:`[Chapter] ${$}`});if(L.length>0)L.push("");L.push(`### ${$}`);L.push("");J++}const Y=k(z.start),Z=z.speaker!==void 0?`Speaker ${z.speaker+1}`:"";V.push({timestamp:Y,speaker:Z,text:z.text});if(z.speakerChange&&L.length>0)L.push("");L.push(`${Y} ${z.text}`)}return{rows:V,plainText:L.join(`
+`)}}function k(q){const K=Math.floor(q/3600),j=Math.floor(q%3600/60),J=Math.floor(q%60);if(K>0)return`${K}:${String(j).padStart(2,"0")}:${String(J).padStart(2,"0")}`;return`${j}:${String(J).padStart(2,"0")}`}const O=30;function B(q){const K=[];let j="",J=0,V=0;const L=()=>{if(j.trim()){K.push({start:J,text:j.trim(),speakerChange:!1});j=""}};for(const z of q){if(j&&z.start-V>M)L();if(j&&z.start-J>O)L();if(!j)J=z.start;j+=(j?" ":"")+z.text;V=z.start;if(y.test(z.text))L()}L();return K}function b(q){const K=[];let j=null,J=-1,V="";for(const z of q){const Y=/^>>/.test(z.text),Z=z.text.replace(/^>>\s*/,"").replace(/^-\s+/,""),$=/,\s*$/.test(V),Q=(y.test(V)||!V)&&!$;if(Y&&Q){if(j)K.push(j);J=(J+1)%2;j={start:z.start,segments:[{start:z.start,text:Z}],speakerChange:!0,speaker:J}}else{if(!j)j={start:z.start,segments:[],speakerChange:!1};j.segments.push({start:z.start,text:Z})}V=Z}if(j)K.push(j);C(K);const L=[];for(const z of K){const Y=z.speaker===void 0?B(z.segments):l(B(z.segments));for(let Z=0;Z<Y.length;Z++)L.push({...Y[Z],speakerChange:Z===0&&z.speakerChange,speaker:z.speaker})}return L}function C(q){const K=/^(mhm|yeah|yes|yep|right|okay|ok|absolutely|sure|exactly|uh-huh|mm-hmm)[.!,]?\s+/i;for(let j=0;j<q.length;j++){const J=q[j];if(J.speaker===void 0||J.segments.length===0)continue;const V=J.segments[0],L=K.exec(V.text);if(!L)continue;if(/,\s*$/.test(L[0]))continue;const z=V.text.slice(L[0].length).trim(),Y=J.segments.slice(1);if(H(z)+Y.reduce((X,w)=>X+H(w.text),0)<30)continue;const $=L[0].trimEnd(),Q=z?[{start:V.start,text:z},...Y]:Y;q.splice(j,1,{start:J.start,segments:[{start:V.start,text:$}],speakerChange:J.speakerChange,speaker:J.speaker},{start:Q[0].start,segments:Q,speakerChange:!0,speaker:J.speaker===0?1:0});j++}}function l(q){if(q.length<=1)return q;const K=[];let j={...q[0]},J=!0;for(let V=1;V<q.length;V++){const L=q[V];if(W(j,L,J)){j.text=`${j.text} ${L.text}`;continue}K.push(j);j={...L};J=!1}K.push(j);return K}function W(q,K,j){const J=H(q.text),V=H(K.text);if(v(q.text,J)||v(K.text,V))return!1;if(j&&J<G)return!1;if(F.test(q.text)||F.test(K.text))return!1;if(J+V>A)return!1;if(K.start-q.start>U)return!1;return!0}function v(q,K){const j=K??H(q);return j>0&&j<=D&&y.test(q)}

package/clis/youtube/transcript.js

@@ -1,166 +1,12 @@
-/**
- * YouTube transcript — extracts caption tracks through the watch player.
- *
- * YouTube timedtext URLs can require player-generated PO tokens. Load the
- * watch page, ask the player captions module to select a track, then reuse the
- * generated json3 timedtext URL before falling back to watch HTML captions.
- *
- * Modes:
- *   --mode grouped (default): sentences merged, speaker detection, chapters
- *   --mode raw: every caption segment as-is with precise timestamps
- */
-import { cli, Strategy } from '@jackwener/opencli/registry';
-import { extractJsonAssignmentFromHtml, parseVideoId, prepareYoutubeApiPage } from './utils.js';
-import { groupTranscriptSegments, formatGroupedTranscript, } from './transcript-group.js';
-import { CommandExecutionError, EmptyResultError } from '@jackwener/opencli/errors';
-
-function unwrapBrowserResult(value) {
-    if (value && typeof value === 'object' && 'session' in value && 'data' in value) {
-        return value.data;
-    }
-    return value;
-}
-
-function normalizeSegmentsPayload(value, source, { allowNull = false } = {}) {
-    const payload = unwrapBrowserResult(value);
-    if (payload == null && allowNull)
-        return null;
-    if (Array.isArray(payload))
-        return payload;
-    if (payload && typeof payload === 'object' && payload.error) {
-        throw new CommandExecutionError(String(payload.error));
-    }
-    throw new CommandExecutionError(`Malformed ${source} payload`);
-}
-
-function parseJson3Segments(text) {
-    let data;
-    try {
-        data = JSON.parse(text);
-    }
-    catch (err) {
-        throw new CommandExecutionError(`Malformed json3 timedtext response: ${err?.message || err}`);
-    }
-    if (!Array.isArray(data?.events)) {
-        throw new CommandExecutionError('Malformed json3 timedtext response: missing events array');
-    }
-    const rows = [];
-    for (const event of data.events) {
-        const startMs = Number(event?.tStartMs || 0);
-        const durMs = Number(event?.dDurationMs || 0);
-        const segs = Array.isArray(event?.segs) ? event.segs : [];
-        const line = segs.map(seg => seg?.utf8 || '').join('').replace(/\s+/g, ' ').trim();
-        if (!line)
-            continue;
-        rows.push({
-            start: startMs / 1000,
-            end: (startMs + durMs) / 1000,
-            text: line,
-        });
-    }
-    return rows;
-}
-
-function timedtextUrlMatchesVideo(url, videoId) {
-    if (!videoId)
-        return true;
-    try {
-        const parsed = new URL(url);
-        return parsed.searchParams.get('v') === videoId;
-    }
-    catch {
-        return false;
-    }
-}
-
-function extractSegmentsFromNetworkCapture(entries, lang, videoId) {
-    const payload = unwrapBrowserResult(entries);
-    if (!Array.isArray(payload) || payload.length === 0)
-        return { segments: [] };
-    const wanted = String(lang || '').toLowerCase();
-    const wantedBase = wanted.split('-')[0];
-    const timedtext = payload
-        .filter((entry) => {
-        const url = String(entry?.url || '');
-        if (!url.includes('/api/timedtext'))
-            return false;
-        if (!url.includes('fmt=json3') || !url.includes('pot='))
-            return false;
-        // Scope to the current video — daemon-shared tabs can retain captured
-        // timedtext entries from prior YouTube SPA navigations that match
-        // the same lang. Use exact query-param equality rather than substring
-        // matching so v=<prefix> cannot match v=<prefix><suffix>.
-        if (!timedtextUrlMatchesVideo(url, videoId))
-            return false;
-        if (!wanted)
-            return true;
-        try {
-            const u = new URL(url);
-            const got = String(u.searchParams.get('lang') || '').toLowerCase();
-            const gotBase = got.split('-')[0];
-            return got === wanted || gotBase === wantedBase || wantedBase === got;
-        }
-        catch {
-            return false;
-        }
-    })
-        .reverse();
-    let malformed = '';
-    for (const entry of timedtext) {
-        const body = typeof entry?.responsePreview === 'string' ? entry.responsePreview : '';
-        if (!body)
-            continue;
-        try {
-            const parsed = parseJson3Segments(body);
-            if (parsed.length > 0)
-                return { segments: parsed };
-        }
-        catch (err) {
-            malformed = err?.message || String(err);
-        }
-    }
-    return malformed ? { error: malformed } : { segments: [] };
-}
-
-cli({
-    site: 'youtube',
-    name: 'transcript',
-    access: 'read',
-    description: 'Get YouTube video transcript/subtitles',
-    domain: 'www.youtube.com',
-    strategy: Strategy.COOKIE,
-    args: [
-        { name: 'url', required: true, positional: true, help: 'YouTube video URL or video ID' },
-        { name: 'lang', required: false, help: 'Language code (e.g. en, zh-Hans). Omit to auto-select' },
-        { name: 'mode', required: false, default: 'grouped', help: 'Output mode: grouped (readable paragraphs) or raw (every segment)' },
-    ],
-    // columns intentionally omitted — raw and grouped modes return different schemas,
-    // so we let the renderer auto-detect columns from the data keys.
-    func: async (page, kwargs) => {
-        const videoId = parseVideoId(kwargs.url);
-        const lang = kwargs.lang || '';
-        const mode = kwargs.mode || 'grouped';
-        const watchUrl = 'https://www.youtube.com/watch?v=' + encodeURIComponent(videoId);
-        const canCapture = typeof page.startNetworkCapture === 'function' && typeof page.readNetworkCapture === 'function';
-        if (canCapture) {
-            try {
-                await page.startNetworkCapture('/api/timedtext');
-            }
-            catch {
-                // Best-effort only. The in-page capture and XML fallback still run.
-            }
-        }
-        await page.goto(watchUrl, { waitUntil: 'none' });
-        await page.wait(3);
-        const playerResult = await page.evaluate(`
+import{cli as T,Strategy as Y}from"@jackwener/opencli/registry";import{extractJsonAssignmentFromHtml as S,parseVideoId as b,prepareYoutubeApiPage as B}from"./utils.js";import{groupTranscriptSegments as M,formatGroupedTranscript as N}from"./transcript-group.js";import{CommandExecutionError as V,EmptyResultError as G}from"@jackwener/opencli/errors";function A(q){if(q&&typeof q==="object"&&"session"in q&&"data"in q)return q.data;return q}function P(q,K,{allowNull:L=!1}={}){const H=A(q);if(H==null&&L)return null;if(Array.isArray(H))return H;if(H&&typeof H==="object"&&H.error)throw new V(String(H.error));throw new V(`Malformed ${K} payload`)}function R(q){let K;try{K=JSON.parse(q)}catch(H){throw new V(`Malformed json3 timedtext response: ${H?.message||H}`)}if(!Array.isArray(K?.events))throw new V("Malformed json3 timedtext response: missing events array");const L=[];for(const H of K.events){const Q=Number(H?.tStartMs||0),X=Number(H?.dDurationMs||0),W=(Array.isArray(H?.segs)?H.segs:[]).map((J)=>J?.utf8||"").join("").replace(/\s+/g," ").trim();if(!W)continue;L.push({start:Q/1000,end:(Q+X)/1000,text:W})}return L}function f(q,K){if(!K)return!0;try{return new URL(q).searchParams.get("v")===K}catch{return!1}}function h(q,K,L){const H=A(q);if(!Array.isArray(H)||H.length===0)return{segments:[]};const Q=String(K||"").toLowerCase(),X=Q.split("-")[0],Z=H.filter((J)=>{const j=String(J?.url||"");if(!j.includes("/api/timedtext"))return!1;if(!j.includes("fmt=json3")||!j.includes("pot="))return!1;if(!f(j,L))return!1;if(!Q)return!0;try{const O=new URL(j),_=String(O.searchParams.get("lang")||"").toLowerCase(),F=_.split("-")[0];return _===Q||F===X||X===_}catch{return!1}}).reverse();let W="";for(const J of Z){const j=typeof J?.responsePreview==="string"?J.responsePreview:"";if(!j)continue;try{const O=R(j);if(O.length>0)return{segments:O}}catch(O){W=O?.message||String(O)}}return W?{error:W}:{segments:[]}}T({site:"youtube",name:"transcript",access:"read",description:"Get YouTube video transcript/subtitles",domain:"www.youtube.com",strategy:Y.COOKIE,args:[{name:"url",required:!0,positional:!0,help:"YouTube video URL or video ID"},{name:"lang",required:!1,help:"Language code (e.g. en, zh-Hans). Omit to auto-select"},{name:"mode",required:!1,default:"grouped",help:"Output mode: grouped (readable paragraphs) or raw (every segment)"}],func:async(q,K)=>{const L=b(K.url),H=K.lang||"",Q=K.mode||"grouped",X="https://www.youtube.com/watch?v="+encodeURIComponent(L),Z=typeof q.startNetworkCapture==="function"&&typeof q.readNetworkCapture==="function";if(Z)try{await q.startNetworkCapture("/api/timedtext")}catch{}await q.goto(X,{waitUntil:"none"});await q.wait(3);const W=await q.evaluate(`
       (async () => {
-        const langPref = ${JSON.stringify(lang)};
+        const langPref = ${JSON.stringify(H)};
         // Scope all timedtext URL matching to the current video. YouTube is an
         // SPA, so watch→watch navigations preserve performance.getEntriesByType
         // entries from prior videos. Without this check a stale same-language
         // URL can be picked up by the polling loop before the current video's
         // fetch hook fires, leaking the predecessor's captions.
-        const targetVideoId = ${JSON.stringify(videoId)};
+        const targetVideoId = ${JSON.stringify(L)};
         const sleep = (ms) => new Promise(resolve => setTimeout(resolve, ms));
 
         function textFromJson3Event(event) {
@@ -375,33 +221,11 @@ cli({
           if (OriginalXHR) globalThis.XMLHttpRequest = OriginalXHR;
         }
       })()
-    `);
-        let segments = normalizeSegmentsPayload(playerResult, 'player caption extraction', { allowNull: true });
-        if (!segments && canCapture) {
-            try {
-                const captured = extractSegmentsFromNetworkCapture(await page.readNetworkCapture(), lang, videoId);
-                if (captured.error) {
-                    throw new CommandExecutionError(captured.error);
-                }
-                if (captured.segments.length > 0) {
-                    segments = captured.segments;
-                }
-            }
-            catch (err) {
-                if (err instanceof CommandExecutionError)
-                    throw err;
-                // Keep existing fallback path when capture is unavailable.
-            }
-        }
-        if (!segments) {
-            await prepareYoutubeApiPage(page);
-        }
-        // Fallback: get caption track URL from watch page HTML
-        const captionData = segments ? null : unwrapBrowserResult(await page.evaluate(`
+    `);let J=P(W,"player caption extraction",{allowNull:!0});if(!J&&Z)try{const z=h(await q.readNetworkCapture(),H,L);if(z.error)throw new V(z.error);if(z.segments.length>0)J=z.segments}catch(z){if(z instanceof V)throw z}if(!J)await B(q);const j=J?null:A(await q.evaluate(`
       (async () => {
-        const extractJsonAssignmentFromHtml = ${extractJsonAssignmentFromHtml.toString()};
+        const extractJsonAssignmentFromHtml = ${S.toString()};
 
-        const watchResp = await fetch('/watch?v=' + encodeURIComponent(${JSON.stringify(videoId)}), {
+        const watchResp = await fetch('/watch?v=' + encodeURIComponent(${JSON.stringify(L)}), {
           credentials: 'include',
         });
         if (!watchResp.ok) return { error: 'Watch HTML returned HTTP ' + watchResp.status };
@@ -418,7 +242,7 @@ cli({
         const tracks = renderer.captionTracks;
         const available = tracks.map(t => t.languageCode + (t.kind === 'asr' ? ' (auto)' : ''));
 
-        const langPref = ${JSON.stringify(lang)};
+        const langPref = ${JSON.stringify(H)};
         let track = null;
         if (langPref) {
           track = tracks.find(t => t.languageCode === langPref)
@@ -438,37 +262,7 @@ cli({
           langPrefixMatched: !!(langPref && track.languageCode !== langPref && track.languageCode.startsWith(langPref))
         };
       })()
-    `));
-        if (!segments && (!captionData || typeof captionData !== 'object' || Array.isArray(captionData))) {
-            throw new CommandExecutionError(`Failed to get caption info: ${typeof captionData === 'string' ? captionData : 'malformed response'}`);
-        }
-        if (captionData?.error) {
-            const msg = `${captionData.error}${captionData.available ? ' (available: ' + captionData.available.join(', ') + ')' : ''}`;
-            // "No captions available" 是合法 empty 数据条件（作者没开字幕 + YT 没自动生成），
-            // 与 bilibili subtitle 的 EmptyResultError 同模式。下游应按 code EMPTY_RESULT 跳过
-            // 重试和 softFail 计数。其它 error（HTTP / parse / 短暂空响应）仍按 fetch 失败抛。
-            if (captionData.error === 'No captions available for this video') {
-                throw new EmptyResultError('youtube transcript', '该视频没有字幕（作者未开启 + 无自动字幕）。');
-            }
-            throw new CommandExecutionError(msg);
-        }
-        if (!segments && typeof captionData?.captionUrl !== 'string') {
-            throw new CommandExecutionError('Malformed caption info payload');
-        }
-        // Warn if --lang was specified but not matched
-        if (captionData?.requestedLang && !captionData.langMatched && !captionData.langPrefixMatched) {
-            console.error(`Warning: --lang "${captionData.requestedLang}" not found. Using "${captionData.language}" instead. Available: ${captionData.available.join(', ')}`);
-        }
-        // Step 2: Fetch caption XML and parse segments
-        // Ensure caption URL requests srv3 XML format — YouTube may return empty
-        // responses when no explicit format is specified.
-        if (!segments) {
-            const originalCaptionUrl = captionData.captionUrl;
-            let captionUrl = originalCaptionUrl;
-            if (!/[&?]fmt=/.test(originalCaptionUrl)) {
-                captionUrl = originalCaptionUrl + (originalCaptionUrl.includes('?') ? '&' : '?') + 'fmt=srv3';
-            }
-            segments = normalizeSegmentsPayload(await page.evaluate(`
+    `));if(!J&&(!j||typeof j!=="object"||Array.isArray(j)))throw new V(`Failed to get caption info: ${typeof j==="string"?j:"malformed response"}`);if(j?.error){const z=`${j.error}${j.available?" (available: "+j.available.join(", ")+")":""}`;if(j.error==="No captions available for this video")throw new G("youtube transcript","该视频没有字幕（作者未开启 + 无自动字幕）。");throw new V(z)}if(!J&&typeof j?.captionUrl!=="string")throw new V("Malformed caption info payload");if(j?.requestedLang&&!j.langMatched&&!j.langPrefixMatched)console.error(`Warning: --lang "${j.requestedLang}" not found. Using "${j.language}" instead. Available: ${j.available.join(", ")}`);if(!J){const z=j.captionUrl;let $=z;if(!/[&?]fmt=/.test(z))$=z+(z.includes("?")?"&":"?")+"fmt=srv3";J=P(await q.evaluate(`
       (async () => {
         async function fetchCaptionXml(url) {
           const resp = await fetch(url);
@@ -476,8 +270,8 @@ cli({
           return { xml: await resp.text() || '' };
         }
 
-        const primaryUrl = ${JSON.stringify(captionUrl)};
-        const originalUrl = ${JSON.stringify(originalCaptionUrl)};
+        const primaryUrl = ${JSON.stringify($)};
+        const originalUrl = ${JSON.stringify(z)};
         let result = await fetchCaptionXml(primaryUrl);
         if (result.error) return result;
 
@@ -556,16 +350,7 @@ cli({
 
         return results;
       })()
-    `), 'caption XML extraction');
-        }
-        if (segments.length === 0) {
-            throw new EmptyResultError('youtube transcript');
-        }
-        // Step 3: Fetch chapters (for grouped mode)
-        let chapters = [];
-        if (mode === 'grouped') {
-            try {
-                const chapterData = unwrapBrowserResult(await page.evaluate(`
+    `),"caption XML extraction")}if(J.length===0)throw new G("youtube transcript");let O=[];if(Q==="grouped")try{const z=A(await q.evaluate(`
           (async () => {
             const cfg = window.ytcfg?.data_ || {};
             const apiKey = cfg.INNERTUBE_API_KEY;
@@ -577,7 +362,7 @@ cli({
               headers: { 'Content-Type': 'application/json' },
               body: JSON.stringify({
                 context: { client: { clientName: 'WEB', clientVersion: '2.20240101.00.00' } },
-                videoId: ${JSON.stringify(videoId)}
+                videoId: ${JSON.stringify(L)}
               })
             });
             if (!resp.ok) return [];
@@ -629,28 +414,4 @@ cli({
             }
             return chapters;
           })()
-        `));
-                if (Array.isArray(chapterData)) {
-                    chapters = chapterData;
-                }
-            }
-            catch {
-                // Chapters are optional — proceed without them
-            }
-        }
-        // Step 4: Format output based on mode
-        if (mode === 'raw') {
-            // Precise timestamps in seconds with decimals, matching bilibili/subtitle format
-            return segments.map((seg, i) => ({
-                index: i + 1,
-                start: Number(seg.start).toFixed(2) + 's',
-                end: Number(seg.end).toFixed(2) + 's',
-                text: seg.text,
-            }));
-        }
-        // Grouped mode: merge sentences, detect speakers, insert chapters
-        const grouped = groupTranscriptSegments(segments.map(s => ({ start: s.start, text: s.text })));
-        const { rows } = formatGroupedTranscript(grouped, chapters);
-        return rows;
-    },
-});
+        `));if(Array.isArray(z))O=z}catch{}if(Q==="raw")return J.map((z,$)=>({index:$+1,start:Number(z.start).toFixed(2)+"s",end:Number(z.end).toFixed(2)+"s",text:z.text}));const _=M(J.map((z)=>({start:z.start,text:z.text}))),{rows:F}=N(_,O);return F}});