@tacone/prosey 0.5.0 → 0.6.0

package/bin/prosey

@@ -103291,7 +103291,7 @@ function openInBrowser(htmlPath) {
 // package.json
 var package_default = {
   name: "@tacone/prosey",
-  version: "0.5.0",
+  version: "0.6.0",
   description: "Download YouTube video transcripts from the CLI",
   module: "src/index.ts",
   type: "module",
@@ -122284,10 +122284,12 @@ try {
     const dir2 = cacheDir(videoId, cacheOpts2);
     let segments2 = null;
     let summary = null;
+    let cachedInfo = null;
     startTimer();
     if (!noCache) {
       const cachedSegments = await readCache(dir2, "transcript.json");
       const cachedSummary = await readCache(dir2, "summary.md");
+      cachedInfo = await readCache(dir2, "info.json");
       if (cachedSegments && cachedSummary) {
         info("Transcript cached");
         debug("Cache hit:", dir2);
@@ -122299,23 +122301,83 @@ try {
     } else {
       debug("Cache skipped (--no-cache)");
     }
-    if (!segments2) {
-      info("Fetching transcript...");
-      segments2 = lang ? await fetchTranscript(videoId, { lang }) : await fetchTranscript(videoId);
-      info(`Transcript: ${segments2.length} segments`);
-      await writeCache(dir2, "transcript.json", JSON.stringify(segments2));
-      debug("Cache written: transcript.json");
-    }
     const prompt2 = resolveSummarizePrompt(config) ?? "";
     if (!prompt2) {
       console.error("Error: no prompt configured. Set a prompt in the [summarize] section of your config.");
       exitProcess(1);
     }
-    const transcriptText = toText(segments2, !noDecode);
+    let structuredContent;
+    if (!segments2) {
+      info("Fetching transcript...");
+      const opts = lang ? { lang, videoDetails: true } : { videoDetails: true };
+      const result = await fetchTranscript(videoId, opts);
+      segments2 = result.segments;
+      const infoJson = JSON.stringify({
+        title: result.videoDetails.title,
+        channel: result.videoDetails.author,
+        description: result.videoDetails.description
+      });
+      cachedInfo = infoJson;
+      const chapterValue = formatChaptersAsJson(extractChapters(result.videoDetails.description));
+      const truncatedInfo = JSON.stringify({
+        title: result.videoDetails.title,
+        channel: result.videoDetails.author,
+        description: result.videoDetails.description.slice(0, 1000)
+      });
+      const transcriptText = toText(segments2, !noDecode);
+      structuredContent = `INFO:
+${truncatedInfo}
+
+TEXT:
+${transcriptText}`;
+      if (dryRun) {
+        await outputText(`${prompt2}
+
+${structuredContent}
+`);
+        exitProcess(0);
+      }
+      info(`Transcript: ${segments2.length} segments`);
+      await writeCache(dir2, "transcript.json", JSON.stringify(segments2));
+      await writeCache(dir2, "info.json", infoJson);
+      await writeCache(dir2, "chapters.json", chapterValue);
+      debug("Cache written: transcript.json, info.json, chapters.json");
+    } else {
+      let chapterValue;
+      if (!cachedInfo) {
+        debug("Cache missing info.json, re-fetching video details");
+        const fallbackOpts = lang ? { lang, videoDetails: true } : { videoDetails: true };
+        const fallbackResult = await fetchTranscript(videoId, fallbackOpts);
+        cachedInfo = JSON.stringify({
+          title: fallbackResult.videoDetails.title,
+          channel: fallbackResult.videoDetails.author,
+          description: fallbackResult.videoDetails.description
+        });
+        await writeCache(dir2, "info.json", cachedInfo);
+        chapterValue = formatChaptersAsJson(extractChapters(fallbackResult.videoDetails.description));
+        await writeCache(dir2, "chapters.json", chapterValue);
+        debug("Cache written: info.json, chapters.json");
+      } else {
+        const cachedChapters = await readCache(dir2, "chapters.json");
+        chapterValue = cachedChapters ?? "not available";
+      }
+      const transcriptText = toText(segments2, !noDecode);
+      const cachedInfoObj = JSON.parse(cachedInfo);
+      const truncatedInfo = JSON.stringify({
+        title: cachedInfoObj.title,
+        channel: cachedInfoObj.channel,
+        description: cachedInfoObj.description.slice(0, 1000)
+      });
+      structuredContent = `INFO:
+${truncatedInfo}
+
+TEXT:
+${transcriptText}`;
+    }
     if (dryRun) {
       await outputText(`${prompt2}
 
-${transcriptText}
+${structuredContent}
 `);
       exitProcess(0);
     }
@@ -122324,7 +122386,7 @@ ${transcriptText}
       summary = await summarize({
         prompt: prompt2,
         command: sumCmd,
-        transcript: transcriptText,
+        transcript: structuredContent,
         cwd: dir2
       });
       info("Summary ready");

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tacone/prosey",
-  "version": "0.5.0",
+  "version": "0.6.0",
   "description": "Download YouTube video transcripts from the CLI",
   "module": "src/index.ts",
   "type": "module",

package/src/default-config.toml

@@ -29,7 +29,23 @@ command = "OPENCODE_PERMISSION='{\"read\":\"allow\",\"write\":\"deny\",\"edit\":
 # Customize this to change how transcripts are summarized.
 
 prompt = """
-Write a comprehensive summary of the following transcription.
+Write a comprehensive summary of the following video transcription:
+
+Guidelines:
+- use the INFO section data to correct spellings in the text (e.g. Pi vs Pie)
+- use the INFO title value as the main title (h1 header)
+- use the TIMESTAMPS only to understand the structure, don't output time markers
+- content length: aim to be the same range as the original text or more.
+- section titles should be h2 headers
+
+Content considerations:
+
+- findings should be presented in lists or table
+- flows should be presented in ordered lists
+- elaborations are better in prose
+- when content mixes types within a section, lead with prose and break out lists only for discrete items
+
+Only output the summary text without adding anything.
 """
 
 # Command override for summarize. Uncomment to use a different command

package/src/index.ts

@@ -443,12 +443,14 @@ try {
     const dir = cacheDir(videoId, cacheOpts);
     let segments: TranscriptSegment[] | null = null;
     let summary: string | null = null;
+    let cachedInfo: string | null = null;
 
     startTimer();
 
     if (!noCache) {
       const cachedSegments = await readCache(dir, "transcript.json");
       const cachedSummary = await readCache(dir, "summary.md");
+      cachedInfo = await readCache(dir, "info.json");
       if (cachedSegments && cachedSummary) {
         info("Transcript cached");
         debug("Cache hit:", dir);
@@ -461,14 +463,6 @@ try {
       debug("Cache skipped (--no-cache)");
     }
 
-    if (!segments) {
-      info("Fetching transcript...");
-      segments = lang ? await fetchTranscript(videoId, { lang }) : await fetchTranscript(videoId);
-      info(`Transcript: ${segments.length} segments`);
-      await writeCache(dir, "transcript.json", JSON.stringify(segments));
-      debug("Cache written: transcript.json");
-    }
-
     const prompt = resolveSummarizePrompt(config) ?? "";
     if (!prompt) {
       console.error(
@@ -476,10 +470,80 @@ try {
       );
       exitProcess(1);
     }
-    const transcriptText = toText(segments, !noDecode);
+
+    let structuredContent: string;
+
+    if (!segments) {
+      info("Fetching transcript...");
+      const opts = lang ? { lang, videoDetails: true as const } : { videoDetails: true as const };
+      const result = (await fetchTranscript(videoId, opts)) as {
+        videoDetails: VideoDetails;
+        segments: TranscriptSegment[];
+      };
+      segments = result.segments;
+      const infoJson = JSON.stringify({
+        title: result.videoDetails.title,
+        channel: result.videoDetails.author,
+        description: result.videoDetails.description,
+      });
+      cachedInfo = infoJson;
+      const chapterValue = formatChaptersAsJson(extractChapters(result.videoDetails.description));
+      const truncatedInfo = JSON.stringify({
+        title: result.videoDetails.title,
+        channel: result.videoDetails.author,
+        description: result.videoDetails.description.slice(0, 1000),
+      });
+      const transcriptText = toText(segments, !noDecode);
+      structuredContent = `INFO:\n${truncatedInfo}\n\nTEXT:\n${transcriptText}`;
+
+      if (dryRun) {
+        await outputText(`${prompt}\n\n${structuredContent}\n`);
+        exitProcess(0);
+      }
+
+      info(`Transcript: ${segments.length} segments`);
+      await writeCache(dir, "transcript.json", JSON.stringify(segments));
+      await writeCache(dir, "info.json", infoJson);
+      await writeCache(dir, "chapters.json", chapterValue);
+      debug("Cache written: transcript.json, info.json, chapters.json");
+    } else {
+      let chapterValue: string;
+      if (!cachedInfo) {
+        debug("Cache missing info.json, re-fetching video details");
+        const fallbackOpts = lang
+          ? { lang, videoDetails: true as const }
+          : { videoDetails: true as const };
+        const fallbackResult = (await fetchTranscript(videoId, fallbackOpts)) as {
+          videoDetails: VideoDetails;
+          segments: TranscriptSegment[];
+        };
+        cachedInfo = JSON.stringify({
+          title: fallbackResult.videoDetails.title,
+          channel: fallbackResult.videoDetails.author,
+          description: fallbackResult.videoDetails.description,
+        });
+        await writeCache(dir, "info.json", cachedInfo);
+        chapterValue = formatChaptersAsJson(
+          extractChapters(fallbackResult.videoDetails.description),
+        );
+        await writeCache(dir, "chapters.json", chapterValue);
+        debug("Cache written: info.json, chapters.json");
+      } else {
+        const cachedChapters = await readCache(dir, "chapters.json");
+        chapterValue = cachedChapters ?? "not available";
+      }
+      const transcriptText = toText(segments, !noDecode);
+      const cachedInfoObj = JSON.parse(cachedInfo);
+      const truncatedInfo = JSON.stringify({
+        title: cachedInfoObj.title,
+        channel: cachedInfoObj.channel,
+        description: cachedInfoObj.description.slice(0, 1000),
+      });
+      structuredContent = `INFO:\n${truncatedInfo}\n\nTEXT:\n${transcriptText}`;
+    }
 
     if (dryRun) {
-      await outputText(`${prompt}\n\n${transcriptText}\n`);
+      await outputText(`${prompt}\n\n${structuredContent}\n`);
       exitProcess(0);
     }
 
@@ -488,7 +552,7 @@ try {
       summary = await summarize({
         prompt,
         command: sumCmd,
-        transcript: transcriptText,
+        transcript: structuredContent,
         cwd: dir,
       });
       info("Summary ready");