summd 0.1.8 → 0.1.10

package/dist/url-to-md.js

@@ -78,79 +78,93 @@ async function fetchYouTubeTitle(url, videoId) {
     catch { }
     return `YouTube: ${videoId}`;
 }
-// Browser cookie sources — tried in order, first one that succeeds is used.
-// Cookies bypass bot-detection and geo-blocks; required on most IPs.
+// Browser cookie sources — tried in order, first working one is reused for all runs.
 const BROWSERS = ['chrome', 'chromium', 'firefox', 'safari', 'edge'];
-// yt-dlp — uses browser cookies by default; falls back to no-cookies for public videos.
-// --sub-langs all: accept whatever subtitle language the video provides, no preference.
+// Subtitle strategies — tried in order, stop at first that produces output.
+// Each downloads only the requested languages, not everything.
+const SUB_STRATEGIES = [
+    // 1. Original-language ASR: yt-dlp 'orig' matches the video's primary language
+    ['--write-auto-subs', '--sub-langs', 'orig'],
+    // 2. English manual + ASR
+    ['--write-subs', '--write-auto-subs', '--sub-langs', 'en'],
+    // 3. Most-spoken languages by global user count (covers remaining cases)
+    ['--write-subs', '--write-auto-subs', '--sub-langs', 'zh-Hans,zh,ja,ko,fr,de,es,pt,hi,ar,ru'],
+];
 async function ytDlpTranscript(url, videoId) {
     const dir = tmpdir();
     const outTemplate = join(dir, videoId);
-    const run = (cookieArgs) => execFileAsync('yt-dlp', [
-        '--write-subs',
-        '--write-auto-subs',
-        '--sub-langs', 'all', // accept any language — no hardcoded preference
-        '--sub-format', 'json3',
-        '--skip-download',
-        '--quiet',
-        '--no-progress',
-        ...cookieArgs,
-        '-o', outTemplate,
-        url,
-    ], { timeout: 30_000 });
-    // Try with browser cookies first (handles bot-detection, auth, geo-block).
-    // Fall back to no-cookies only if no browser is available.
-    let ran = false;
-    for (const browser of BROWSERS) {
+    const run = async (subArgs, cookieArgs) => {
         try {
-            await run(['--cookies-from-browser', browser]);
-            ran = true;
-            break;
+            await execFileAsync('yt-dlp', [
+                '--sub-format', 'json3',
+                '--skip-download', '--quiet', '--no-progress',
+                ...subArgs, ...cookieArgs,
+                '-o', outTemplate, url,
+            ], { timeout: 30_000 });
+            return 'ok';
         }
         catch (e) {
             const err = e;
             if (err.code === 'ENOENT')
-                return { transcript: null, reason: 'not-installed' };
+                return 'not-installed';
             const stderr = err.stderr ?? '';
-            // Cookie access failed (browser not installed / locked DB) — try next browser
-            if (stderr.includes('browser') || stderr.includes('cookie') || stderr.includes('Could not find'))
-                continue;
-            // yt-dlp ran but failed for another reason (video unavailable, no captions…)
-            ran = true;
-            break;
+            if (stderr.includes('Could not find') || stderr.includes('cookies from browser'))
+                return 'cookie-error';
+            return 'other-error';
         }
-    }
-    // No browser found with usable cookies — try without (works for fully public videos)
-    if (!ran) {
+    };
+    const readAndClean = () => {
         try {
-            await run([]);
-        }
-        catch (e) {
-            const err = e;
-            if (err.code === 'ENOENT')
-                return { transcript: null, reason: 'not-installed' };
-            return { transcript: null, reason: 'no-transcript' };
-        }
-    }
-    // Read the first valid output file; clean up all temp files regardless
-    try {
-        const files = readdirSync(dir).filter(f => f.startsWith(videoId) && f.endsWith('.json3'));
-        for (const file of files) {
-            const filePath = join(dir, file);
-            try {
-                const text = parseTimedText(JSON.parse(readFileSync(filePath, 'utf8')));
-                if (text)
-                    return { transcript: text };
-            }
-            finally {
+            const files = readdirSync(dir).filter(f => f.startsWith(videoId) && f.endsWith('.json3'));
+            let transcript = null;
+            for (const file of files) {
+                const filePath = join(dir, file);
                 try {
-                    unlinkSync(filePath);
+                    if (!transcript) {
+                        const text = parseTimedText(JSON.parse(readFileSync(filePath, 'utf8')));
+                        if (text)
+                            transcript = text;
+                    }
+                }
+                finally {
+                    try {
+                        unlinkSync(filePath);
+                    }
+                    catch { }
                 }
-                catch { }
             }
+            return transcript;
+        }
+        catch {
+            return null;
         }
+    };
+    // Phase 1: find a working browser cookie source using the first subtitle strategy.
+    // The result tells us both which browser works AND whether strategy 1 produced output.
+    let cookieArgs = [];
+    for (const browser of BROWSERS) {
+        const outcome = await run(SUB_STRATEGIES[0], ['--cookies-from-browser', browser]);
+        if (outcome === 'not-installed')
+            return { transcript: null, reason: 'not-installed' };
+        if (outcome === 'cookie-error')
+            continue; // browser not available, try next
+        cookieArgs = ['--cookies-from-browser', browser];
+        break;
+    }
+    // cookieArgs is empty if no browser found — fall back to cookie-less requests
+    // Check if strategy 1 produced output
+    const t1 = readAndClean();
+    if (t1)
+        return { transcript: t1 };
+    // Phase 2: try remaining subtitle strategies with the confirmed cookie source
+    for (const subArgs of SUB_STRATEGIES.slice(1)) {
+        const outcome = await run(subArgs, cookieArgs);
+        if (outcome === 'not-installed')
+            return { transcript: null, reason: 'not-installed' };
+        const t = readAndClean();
+        if (t)
+            return { transcript: t };
     }
-    catch { }
     return { transcript: null, reason: 'no-transcript' };
 }
 // json3 parser — shared algorithm with browser extension

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "summd",
-  "version": "0.1.8",
+  "version": "0.1.10",
   "description": "CLI for sum.md — Sum to anything.",
   "license": "MIT",
   "bin": {