tuna-agent 0.1.120 → 0.1.122

package/dist/daemon/analyze-video-handler.d.ts

@@ -26,6 +26,6 @@ export interface AnalyzeVideoResult {
     isError?: boolean;
     error?: string;
 }
-export declare function analyzeVideo(url: string): Promise<AnalyzeVideoResult>;
-/** Extension task handler — wraps analyzeVideo with WS response */
+export declare function analyzeVideo(url: string, onProgress?: (step: string) => void): Promise<AnalyzeVideoResult>;
+/** Extension task handler — wraps analyzeVideo with WS response + progress */
 export declare function handleAnalyzeVideo(ws: AgentWebSocketClient, code: string, taskId: string, url: string): Promise<void>;

package/dist/daemon/analyze-video-handler.js

@@ -97,7 +97,8 @@ async function visionDescribe(frameB64, voiceoverText) {
     const data = await res.json();
     return data.choices?.[0]?.message?.content?.trim() || '';
 }
-export async function analyzeVideo(url) {
+export async function analyzeVideo(url, onProgress) {
+    const progress = onProgress || (() => { });
     const tmpDir = path.join(os.tmpdir(), 'tuna-analyze-' + crypto.randomBytes(6).toString('hex'));
     await fs.mkdir(tmpDir, { recursive: true });
     const videoPath = path.join(tmpDir, 'video.mp4');
@@ -105,8 +106,10 @@ export async function analyzeVideo(url) {
     const framesDir = path.join(tmpDir, 'frames');
     await fs.mkdir(framesDir, { recursive: true });
     try {
+        progress('Đang tải video...');
         console.log('[analyze_video] Downloading:', url);
         await run(YT_DLP, ['-f', 'best[height<=720]/best', '-o', videoPath, '--no-playlist', '--quiet', url]);
+        progress('Đang tách audio...');
         console.log('[analyze_video] Extracting audio');
         await run(FFMPEG, ['-y', '-i', videoPath, '-vn', '-ar', '16000', '-ac', '1', '-b:a', '64k', audioPath, '-loglevel', 'error']);
         console.log('[analyze_video] Probing duration');
@@ -119,35 +122,72 @@ export async function analyzeVideo(url) {
                 return 0;
             }
         })();
+        progress('Đang transcribe bằng Whisper...');
         console.log('[analyze_video] Transcribing via Whisper');
         const rawTranscript = await whisperTranscribe(audioPath);
+        progress('Đang sửa lỗi transcript...');
         console.log('[analyze_video] AI correcting transcript');
         const correctedText = await correctTranscript(rawTranscript.text, rawTranscript.language);
         const transcript = { ...rawTranscript, text: correctedText };
         const segments = transcript.segments || [];
-        console.log('[analyze_video] Extracting', segments.length, 'frames (1 per segment)');
+        const sceneSlots = [];
+        const SILENCE_THRESHOLD = 5; // seconds — gaps longer than this become their own scene
+        const MAX_SCENES = 30;
+        if (segments.length > 0) {
+            // Add silence scene before first segment if gap > threshold
+            if (segments[0].start > SILENCE_THRESHOLD) {
+                sceneSlots.push({ start: 0, end: segments[0].start, voiceover: '' });
+            }
+            for (let i = 0; i < segments.length; i++) {
+                const seg = segments[i];
+                sceneSlots.push({ start: seg.start, end: seg.end, voiceover: seg.text?.trim() || '' });
+                // Add silence scene between segments if gap > threshold
+                if (i < segments.length - 1) {
+                    const gap = segments[i + 1].start - seg.end;
+                    if (gap > SILENCE_THRESHOLD) {
+                        sceneSlots.push({ start: seg.end, end: segments[i + 1].start, voiceover: '' });
+                    }
+                }
+            }
+            // Add silence scene after last segment if gap > threshold
+            const lastEnd = segments[segments.length - 1].end;
+            if (durationSec - lastEnd > SILENCE_THRESHOLD) {
+                sceneSlots.push({ start: lastEnd, end: durationSec, voiceover: '' });
+            }
+        }
+        else {
+            // No transcript — split into scenes every 8s (Veo3 clip length)
+            const interval = 8;
+            for (let t = 0; t < durationSec; t += interval) {
+                sceneSlots.push({ start: t, end: Math.min(t + interval, durationSec), voiceover: '' });
+            }
+        }
+        // Cap max scenes
+        const finalSlots = sceneSlots.slice(0, MAX_SCENES);
+        progress(`Đang cắt ${finalSlots.length} frames và phân tích...`);
+        console.log('[analyze_video] Building', finalSlots.length, 'scenes (segments:', segments.length, ', duration:', durationSec, 's)');
         const scenes = [];
-        for (let i = 0; i < segments.length; i++) {
-            const seg = segments[i];
-            const midpoint = (seg.start + seg.end) / 2;
-            const framePath = path.join(framesDir, `seg-${String(i).padStart(3, '0')}.jpg`);
+        for (let i = 0; i < finalSlots.length; i++) {
+            const slot = finalSlots[i];
+            const midpoint = (slot.start + slot.end) / 2;
+            const framePath = path.join(framesDir, `scene-${String(i).padStart(3, '0')}.jpg`);
             try {
+                progress(`Đang phân tích scene ${i + 1}/${finalSlots.length}...`);
                 await run(FFMPEG, ['-y', '-ss', String(midpoint), '-i', videoPath, '-vframes', '1', '-vf', 'scale=640:-1', '-q:v', '5', framePath, '-loglevel', 'error']);
                 const buf = await fs.readFile(framePath);
-                const voiceover = seg.text?.trim() || '';
-                const visual_description = await visionDescribe(buf.toString('base64'), voiceover);
+                const visual_description = await visionDescribe(buf.toString('base64'), slot.voiceover);
                 scenes.push({
                     scene_number: i + 1,
-                    timestamp_start: Math.round(seg.start * 10) / 10,
-                    timestamp_end: Math.round(seg.end * 10) / 10,
+                    timestamp_start: Math.round(slot.start * 10) / 10,
+                    timestamp_end: Math.round(slot.end * 10) / 10,
                     thumbnail_base64: buf.toString('base64'),
-                    voiceover,
+                    voiceover: slot.voiceover,
                     visual_description,
                 });
             }
             catch (err) {
                 const msg = err instanceof Error ? err.message : String(err);
-                console.warn('[analyze_video] Frame extract failed for segment', i, msg);
+                console.warn('[analyze_video] Frame extract failed for scene', i, msg);
             }
         }
         return {
@@ -165,11 +205,13 @@ export async function analyzeVideo(url) {
         catch { /* ignore */ }
     }
 }
-/** Extension task handler — wraps analyzeVideo with WS response */
+/** Extension task handler — wraps analyzeVideo with WS response + progress */
 export async function handleAnalyzeVideo(ws, code, taskId, url) {
     console.log(`[analyze_video] Starting for ${url}`);
     try {
-        const result = await analyzeVideo(url);
+        const result = await analyzeVideo(url, (step) => {
+            ws.sendExtensionStream(code, taskId, step);
+        });
         ws.sendExtensionDone(code, taskId, { ...result, isError: false });
     }
     catch (err) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "tuna-agent",
-  "version": "0.1.120",
+  "version": "0.1.122",
   "description": "Tuna Agent - Run AI coding tasks on your machine",
   "bin": {
     "tuna-agent": "dist/cli/index.js"