v4.3: file-based media storage + manual VACUUM maintenance

config.py

@@ -16,6 +16,11 @@ STATIC_URL_PATH = '/static'
 # Dev-এ env না থাকলে app.py-র পাশে রাখবে
 DATABASE = os.getenv('DATABASE', os.path.join(BASE_DIR, 'audiobook_maker.db'))
 
+# --- MEDIA STORAGE (v4.3) ---
+# ডেটাবেসের পাশেই media-storage ফোল্ডার রাখি যাতে একই persistent volume-এ থাকে
+_DB_DIR = os.path.dirname(os.path.abspath(DATABASE))
+MEDIA_STORAGE_DIR = os.getenv('MEDIA_STORAGE_DIR', os.path.join(_DB_DIR, 'media-storage'))
+
 # --- FLASK SECRET KEY ---
 SECRET_KEY = os.getenv('SECRET_KEY', 'audiobook-maker-pro-' + str(uuid.uuid4()))
 

db.py

@@ -1,4 +1,4 @@
-# db.py - Database Configuration and Operations (v4.2)
+# db.py - Database Configuration and Operations (v4.3)
 
 import os
 import sqlite3
@@ -32,13 +32,19 @@ def get_db_connection():
 
 
 def init_db():
-    """Initialize database tables. Auto-creates parent directory."""
+    """Initialize database tables. Auto-creates parent directory + media storage."""
     # ফোল্ডার না থাকলে নিজে থেকেই তৈরি করবে (Coolify volume mount-এর জন্য জরুরি)
     db_dir = os.path.dirname(os.path.abspath(DATABASE))
     if db_dir and not os.path.exists(db_dir):
         os.makedirs(db_dir, exist_ok=True)
         print(f"📂 Created data directory: {db_dir}")
 
+    # v4.3: media-storage ফোল্ডার তৈরি
+    from config import MEDIA_STORAGE_DIR
+    if not os.path.exists(MEDIA_STORAGE_DIR):
+        os.makedirs(MEDIA_STORAGE_DIR, exist_ok=True)
+        print(f"📂 Created media storage directory: {MEDIA_STORAGE_DIR}")
+
     with get_db_connection() as conn:
         cursor = conn.cursor()
 
@@ -115,7 +121,7 @@ def init_db():
             )
         ''')
 
-        # v4.2 publishing migrations
+        # v4.2 publishing migrations + v4.3 file-based media path columns
         migrations = [
             ('projects', 'is_published', 'INTEGER DEFAULT 0'),
             ('projects', 'published_at', 'TIMESTAMP'),
@@ -125,6 +131,10 @@ def init_db():
             ('projects', 'author', 'TEXT DEFAULT ""'),
             ('projects', 'category', 'TEXT DEFAULT ""'),
             ('projects', 'view_count', 'INTEGER DEFAULT 0'),
+            # --- v4.3: file-based media path columns ---
+            ('projects', 'thumbnail_path', 'TEXT'),
+            ('markdown_blocks', 'audio_path', 'TEXT'),
+            ('block_images', 'image_path', 'TEXT'),
         ]
 
         for table, column, definition in migrations:
@@ -135,14 +145,42 @@ def init_db():
                 pass
 
         conn.commit()
-        print(f"✅ Database initialized at {DATABASE} (v4.2)")
+        print(f"✅ Database initialized at {DATABASE} (v4.3)")
 
 
 def vacuum_db():
+    """ম্যানুয়াল VACUUM (এখন আর অটোমেটিক চলে না)।"""
     with get_db_connection() as conn:
         conn.execute('VACUUM')
 
 
+def get_db_stats():
+    """
+    ডেটাবেসের সাইজ এবং কত% ফাঁকা (reusable free) স্পেস আছে তা রিটার্ন করে।
+    SQLite-এর freelist_count আর page_count দিয়ে হিসাব করা হয়।
+    """
+    file_size = os.path.getsize(DATABASE) if os.path.exists(DATABASE) else 0
+    with get_db_connection() as conn:
+        cur = conn.cursor()
+        page_count = cur.execute('PRAGMA page_count').fetchone()[0]
+        freelist_count = cur.execute('PRAGMA freelist_count').fetchone()[0]
+        page_size = cur.execute('PRAGMA page_size').fetchone()[0]
+
+    free_bytes = freelist_count * page_size
+    free_percent = round((freelist_count / page_count) * 100, 1) if page_count > 0 else 0.0
+
+    return {
+        'file_size_bytes': file_size,
+        'file_size_mb': round(file_size / (1024 * 1024), 2),
+        'page_count': page_count,
+        'freelist_count': freelist_count,
+        'page_size': page_size,
+        'free_bytes': free_bytes,
+        'free_mb': round(free_bytes / (1024 * 1024), 2),
+        'free_percent': free_percent,
+    }
+
+
 def init_app(app):
     app.teardown_appcontext(close_db)
     init_db()

media_storage.py

@@ -0,0 +1,170 @@
+# media_storage.py - File-based media storage manager (v4.3)
+
+import os
+import base64
+import shutil
+
+from config import MEDIA_STORAGE_DIR
+
+
+def _ensure_dir(path):
+    os.makedirs(path, exist_ok=True)
+
+
+def get_project_dir(project_id):
+    """প্রজেক্টের base ডিরেক্টরি (absolute path) রিটার্ন করে।"""
+    return os.path.join(MEDIA_STORAGE_DIR, f'project_{int(project_id)}')
+
+
+def get_audio_dir(project_id):
+    return os.path.join(get_project_dir(project_id), 'audio')
+
+
+def get_images_dir(project_id):
+    return os.path.join(get_project_dir(project_id), 'images')
+
+
+# ============================================
+# Save operations (base64 string → file)
+# ============================================
+
+def save_audio(project_id, block_id, audio_base64, audio_format='mp3'):
+    """
+    Audio ফাইল সেভ করে relative path রিটার্ন করে।
+    relative path ডেটাবেসে জমা হবে, যেমন: project_5/audio/block_123.mp3
+    """
+    if not audio_base64:
+        return None
+    audio_dir = get_audio_dir(project_id)
+    _ensure_dir(audio_dir)
+    fmt = (audio_format or 'mp3').lower()
+    filename = f'block_{int(block_id)}.{fmt}'
+    abs_path = os.path.join(audio_dir, filename)
+    with open(abs_path, 'wb') as f:
+        f.write(base64.b64decode(audio_base64))
+    return f'project_{int(project_id)}/audio/{filename}'
+
+
+def save_image(project_id, image_id, image_base64, image_format='png'):
+    """Image ফাইল সেভ করে relative path রিটার্ন করে।"""
+    if not image_base64:
+        return None
+    images_dir = get_images_dir(project_id)
+    _ensure_dir(images_dir)
+    fmt = (image_format or 'png').lower()
+    filename = f'img_{int(image_id)}.{fmt}'
+    abs_path = os.path.join(images_dir, filename)
+    with open(abs_path, 'wb') as f:
+        f.write(base64.b64decode(image_base64))
+    return f'project_{int(project_id)}/images/{filename}'
+
+
+def save_thumbnail(project_id, image_bytes, image_format='png'):
+    """Thumbnail সেভ করে relative path রিটার্ন করে (raw bytes নেয়)।"""
+    if not image_bytes:
+        return None
+    proj_dir = get_project_dir(project_id)
+    _ensure_dir(proj_dir)
+    fmt = (image_format or 'png').lower()
+    filename = f'thumbnail.{fmt}'
+    abs_path = os.path.join(proj_dir, filename)
+    # পুরোনো thumbnail অন্য ফরম্যাটে থাকলে মুছে দিই
+    for old_ext in ('png', 'jpeg', 'jpg', 'webp', 'gif'):
+        old = os.path.join(proj_dir, f'thumbnail.{old_ext}')
+        if old_ext != fmt and os.path.exists(old):
+            try:
+                os.remove(old)
+            except OSError:
+                pass
+    with open(abs_path, 'wb') as f:
+        f.write(image_bytes)
+    return f'project_{int(project_id)}/{filename}'
+
+
+# ============================================
+# Read operations (file → bytes / base64)
+# ============================================
+
+def _safe_abs_path(relative_path):
+    """
+    relative path কে absolute path-এ রূপান্তর করে এবং নিশ্চিত করে
+    যে এটা MEDIA_STORAGE_DIR-এর বাইরে যাচ্ছে না (path traversal রোধ)।
+    """
+    if not relative_path:
+        return None
+    base = os.path.realpath(MEDIA_STORAGE_DIR)
+    abs_path = os.path.realpath(os.path.join(base, relative_path))
+    if not abs_path.startswith(base + os.sep) and abs_path != base:
+        return None  # নিরাপত্তা লঙ্ঘন
+    return abs_path
+
+
+def read_file_bytes(relative_path):
+    """ফাইলের raw bytes রিটার্ন করে। না থাকলে None।"""
+    abs_path = _safe_abs_path(relative_path)
+    if not abs_path or not os.path.exists(abs_path):
+        return None
+    with open(abs_path, 'rb') as f:
+        return f.read()
+
+
+def read_file_base64(relative_path):
+    """ফাইল base64 স্ট্রিং হিসেবে রিটার্ন করে। না থাকলে খালি স্ট্রিং।"""
+    data = read_file_bytes(relative_path)
+    if data is None:
+        return ''
+    return base64.b64encode(data).decode('utf-8')
+
+
+def get_safe_abs_path(relative_path):
+    """Flask send_file-এর জন্য নিরাপদ absolute path।"""
+    return _safe_abs_path(relative_path)
+
+
+# ============================================
+# Storage stats (v4.3)
+# ============================================
+
+def get_storage_usage_bytes():
+    """media-storage ফোল্ডারের মোট সাইজ (bytes) রিটার্ন করে।"""
+    total = 0
+    if not os.path.exists(MEDIA_STORAGE_DIR):
+        return 0
+    for root, _, files in os.walk(MEDIA_STORAGE_DIR):
+        for name in files:
+            fp = os.path.join(root, name)
+            try:
+                total += os.path.getsize(fp)
+            except OSError:
+                pass
+    return total
+
+
+# ============================================
+# Delete operations
+# ============================================
+
+def delete_project_media(project_id):
+    """প্রজেক্টের পুরো media ফোল্ডার মুছে দেয় (প্রজেক্ট ডিলিটের সময়)।"""
+    proj_dir = get_project_dir(project_id)
+    if os.path.exists(proj_dir):
+        try:
+            shutil.rmtree(proj_dir)
+            print(f"🗑️ Deleted media folder: {proj_dir}")
+            return True
+        except OSError as e:
+            print(f"⚠️ Failed to delete media folder {proj_dir}: {e}")
+            return False
+    return False
+
+
+def delete_file(relative_path):
+    """একটা নির্দিষ্ট ফাইল মুছে দেয়।"""
+    abs_path = _safe_abs_path(relative_path)
+    if abs_path and os.path.exists(abs_path):
+        try:
+            os.remove(abs_path)
+            return True
+        except OSError:
+            return False
+    return False

migrate_to_files.py

@@ -0,0 +1,96 @@
+# migrate_to_files.py - One-time: base64 in DB → files on disk (v4.2 → v4.3)
+
+import sqlite3
+import base64
+
+from config import DATABASE
+from db import init_db
+from media_storage import save_audio, save_image, save_thumbnail
+
+
+def migrate():
+    # নিশ্চিত করি নতুন কলামগুলো (audio_path, image_path, thumbnail_path) তৈরি আছে
+    print("🔧 Ensuring schema is up to date...")
+    init_db()
+
+    conn = sqlite3.connect(DATABASE)
+    conn.row_factory = sqlite3.Row
+    cur = conn.cursor()
+
+    # --- Audio ---
+    cur.execute('''
+        SELECT mb.id as block_id, mb.audio_data, mb.audio_format, c.project_id
+        FROM markdown_blocks mb JOIN chapters c ON mb.chapter_id = c.id
+        WHERE mb.audio_data IS NOT NULL AND mb.audio_data != ''
+              AND (mb.audio_path IS NULL OR mb.audio_path = '')
+    ''')
+    rows = cur.fetchall()
+    print(f"🔊 Migrating {len(rows)} audio blocks...")
+    done = 0
+    for r in rows:
+        try:
+            rel = save_audio(r['project_id'], r['block_id'], r['audio_data'], r['audio_format'] or 'mp3')
+            if rel:
+                cur.execute("UPDATE markdown_blocks SET audio_path=?, audio_data='' WHERE id=?",
+                            (rel, r['block_id']))
+                done += 1
+        except Exception as e:
+            print(f"  ⚠️ Audio block {r['block_id']} failed: {e}")
+    conn.commit()
+    print(f"   ✅ {done} audio files written")
+
+    # --- Images ---
+    cur.execute('''
+        SELECT bi.id as image_id, bi.image_data, bi.image_format, c.project_id
+        FROM block_images bi
+        JOIN markdown_blocks mb ON bi.block_id = mb.id
+        JOIN chapters c ON mb.chapter_id = c.id
+        WHERE bi.image_data IS NOT NULL AND bi.image_data != ''
+              AND (bi.image_path IS NULL OR bi.image_path = '')
+    ''')
+    rows = cur.fetchall()
+    print(f"🖼️  Migrating {len(rows)} images...")
+    done = 0
+    for r in rows:
+        try:
+            rel = save_image(r['project_id'], r['image_id'], r['image_data'], r['image_format'] or 'png')
+            if rel:
+                cur.execute("UPDATE block_images SET image_path=?, image_data='' WHERE id=?",
+                            (rel, r['image_id']))
+                done += 1
+        except Exception as e:
+            print(f"  ⚠️ Image {r['image_id']} failed: {e}")
+    conn.commit()
+    print(f"   ✅ {done} image files written")
+
+    # --- Thumbnails ---
+    cur.execute('''
+        SELECT id, thumbnail_data, thumbnail_format FROM projects
+        WHERE thumbnail_data IS NOT NULL AND thumbnail_data != ''
+              AND (thumbnail_path IS NULL OR thumbnail_path = '')
+    ''')
+    rows = cur.fetchall()
+    print(f"📕 Migrating {len(rows)} thumbnails...")
+    done = 0
+    for r in rows:
+        try:
+            raw = base64.b64decode(r['thumbnail_data'])
+            rel = save_thumbnail(r['id'], raw, r['thumbnail_format'] or 'png')
+            if rel:
+                cur.execute("UPDATE projects SET thumbnail_path=?, thumbnail_data=NULL WHERE id=?",
+                            (rel, r['id']))
+                done += 1
+        except Exception as e:
+            print(f"  ⚠️ Thumbnail {r['id']} failed: {e}")
+    conn.commit()
+    print(f"   ✅ {done} thumbnails written")
+
+    print("🧹 Running VACUUM (this may take a while on a large DB)...")
+    conn.execute('VACUUM')
+    conn.commit()
+    conn.close()
+    print("✅ Migration complete! Database should now be much smaller.")
+
+
+if __name__ == '__main__':
+    migrate()

routes/generation_routes.py

@@ -1,4 +1,4 @@
-# routes/generation_routes.py - Combined Endpoint with Correct Task Polling
+# routes/generation_routes.py - Combined Endpoint with Correct Task Polling (v4.3)
 
 import json
 import time
@@ -40,8 +40,6 @@ def poll_beam_task(task_id):
     print(f"   URL: {task_url}")
 
     start_time = time.time()
-
-    # প্রথম কয়েকটা attempt এ 404 আসতে পারে — task register হতে delay
     initial_delay = True
 
     while True:
@@ -51,17 +49,14 @@ def poll_beam_task(task_id):
             print(f"❌ Polling timeout after {POLL_MAX_WAIT}s")
             return None, f'Task timed out after {int(POLL_MAX_WAIT)} seconds'
 
-        # প্রথম ২ সেকেন্ড wait করি task register হতে
         if initial_delay and elapsed < 2:
             time.sleep(2)
             initial_delay = False
             continue
 
         try:
-            # ★ Bearer token দিয়ে try
             resp = requests.get(task_url, headers=get_beam_auth_headers(), timeout=30)
 
-            # Bearer fail হলে Basic try করি
             if resp.status_code in (401, 403):
                 print(f"   Bearer auth failed, trying Basic...")
                 basic_headers = {
@@ -73,20 +68,14 @@ def poll_beam_task(task_id):
             print(f"   [{int(elapsed)}s] HTTP {resp.status_code} | Body: {len(resp.text)} chars")
 
             if resp.status_code == 404:
-                # Task এখনও register হয়নি — wait
                 if elapsed < 30:
                     print(f"   Task not found yet, waiting...")
                     time.sleep(POLL_INTERVAL)
                     continue
                 else:
-                    # ৩০ সেকেন্ড পরেও 404 — সমস্যা
                     print(f"❌ Task not found after {int(elapsed)}s")
-
-                    # ★ Debug: response body দেখি
                     print(f"   404 body: {resp.text[:300]}")
 
-                    # ★ Alternative: Beam API base URL ভিন্ন হতে পারে
-                    # কিছু Beam setup এ URL format ভিন্ন
                     alt_urls = [
                         f"https://api.beam.cloud/v2/task/{task_id}/status/",
                         f"https://api.beam.cloud/v2/task/{task_id}",
@@ -121,25 +110,14 @@ def poll_beam_task(task_id):
                 if status in ('COMPLETE', 'COMPLETED', 'SUCCESS'):
                     print(f"✅ Task complete!")
 
-                    # ★ Result বের করা — Beam বিভিন্ন জায়গায় result রাখে
-                    # 1. 'output' key
-                    # 2. 'result' key
-                    # 3. 'outputs' list (file-based)
-                    # 4. response body তেই (endpoint mode)
-
                     actual_result = None
 
-                    # Check 'output' (endpoint mode — function return value)
                     if data.get('output') and isinstance(data['output'], dict):
                         actual_result = data['output']
                         print(f"   Result found in 'output' key")
-
-                    # Check 'result'
                     elif data.get('result') and isinstance(data['result'], dict):
                         actual_result = data['result']
                         print(f"   Result found in 'result' key")
-
-                    # Check if top-level has audio_base64 (unlikely but possible)
                     elif data.get('audio_base64'):
                         actual_result = data
                         print(f"   Result found in top-level data")
@@ -149,16 +127,12 @@ def poll_beam_task(task_id):
                     elif actual_result and actual_result.get('success'):
                         return actual_result, None
 
-                    # ★ Outputs (file-based) — need to download
                     outputs = data.get('outputs', [])
                     if outputs:
                         print(f"   Task has {len(outputs)} output files")
-                        # For our use case, result should be in 'output' not files
-                        # But log it for debug
                         for out in outputs:
                             print(f"   Output: {out.get('name', '?')} → {out.get('url', '?')}")
 
-                    # No usable result found
                     print(f"   ⚠️ Task complete but no audio in response")
                     print(f"   Response keys: {list(data.keys())}")
                     print(f"   Full response (first 500): {json.dumps(data, default=str)[:500]}")
@@ -177,7 +151,7 @@ def poll_beam_task(task_id):
                     return None, f'Task {status.lower()} on Beam. Container may not have started in time.'
 
                 elif status in ('PENDING', 'RUNNING', 'RETRY'):
-                    pass  # Keep polling
+                    pass
 
                 else:
                     print(f"   Unknown status: {status}")
@@ -221,16 +195,12 @@ def call_beam_and_get_result(text, voice='af_heart', speed=1.0):
 
     task_id = response.headers.get('X-Task-Id', '')
 
-    # ========================================
-    # CASE 1: Task ID + empty/no body → Async → Poll
-    # ========================================
+    # CASE 1: Task ID + empty body → Async → Poll
     if task_id and (not response.text or not response.text.strip() or response.headers.get('Content-Length') == '0'):
         print(f"📋 Async mode — Task ID: {task_id}")
         return poll_beam_task(task_id)
 
-    # ========================================
     # CASE 2: Task ID + body
-    # ========================================
     if task_id and response.text and response.text.strip():
         print(f"📋 Task ID: {task_id} + body ({len(response.text)} chars)")
         try:
@@ -238,20 +208,15 @@ def call_beam_and_get_result(text, voice='af_heart', speed=1.0):
             if result.get('success') and result.get('audio_base64'):
                 print(f"✅ Direct sync result")
                 return _extract(result), None
-            # Body isn't the final result — poll
             return poll_beam_task(task_id)
         except Exception:
             return poll_beam_task(task_id)
 
-    # ========================================
     # CASE 3: No task_id + empty body → Error
-    # ========================================
     if not response.text or not response.text.strip():
         return None, 'Empty response from Beam with no task ID'
 
-    # ========================================
     # CASE 4: Synchronous response
-    # ========================================
     if response.status_code != 200:
         try:
             err = response.json().get('error', response.text[:200])
@@ -326,15 +291,26 @@ def generate_audio():
         if source_format != 'mp3':
             audio_base64 = convert_to_mp3(audio_base64, source_format)
 
+        # block_id থাকলে সরাসরি ফাইলে সেভ করি (v4.3)
         if block_id:
+            from media_storage import save_audio
             db = get_db()
             cursor = db.cursor()
             cursor.execute('''
-                UPDATE markdown_blocks 
-                SET audio_data = ?, audio_format = 'mp3', transcription = ?
-                WHERE id = ?
-            ''', (audio_base64, json.dumps(transcription), block_id))
-            db.commit()
+                SELECT c.project_id FROM markdown_blocks mb
+                JOIN chapters c ON mb.chapter_id = c.id
+                WHERE mb.id = ?
+            ''', (block_id,))
+            row = cursor.fetchone()
+            if row:
+                project_id = row['project_id']
+                rel_path = save_audio(project_id, block_id, audio_base64, 'mp3')
+                cursor.execute('''
+                    UPDATE markdown_blocks 
+                    SET audio_path = ?, audio_data = '', audio_format = 'mp3', transcription = ?
+                    WHERE id = ?
+                ''', (rel_path, json.dumps(transcription), block_id))
+                db.commit()
 
         print(f"✅ DONE: audio={len(audio_base64)} bytes, words={len(transcription)}")
         print(f"{'='*60}")
@@ -377,7 +353,7 @@ def generate_chapter_audio():
     cursor = db.cursor()
 
     cursor.execute('''
-        SELECT id, content, tts_text, block_type FROM markdown_blocks 
+        SELECT id, content, tts_text, block_type, chapter_id FROM markdown_blocks 
         WHERE chapter_id = ? ORDER BY block_order
     ''', (chapter_id,))
     blocks = cursor.fetchall()
@@ -385,6 +361,11 @@ def generate_chapter_audio():
     if not blocks:
         return jsonify({'error': 'No blocks found'}), 404
 
+    # project_id বের করি (ফাইল সেভের জন্য)
+    cursor.execute('SELECT project_id FROM chapters WHERE id = ?', (chapter_id,))
+    ch_row = cursor.fetchone()
+    project_id = ch_row['project_id'] if ch_row else None
+
     results = []
     success_count = 0
     error_count = 0
@@ -394,6 +375,8 @@ def generate_chapter_audio():
     print(f"📖 CHAPTER: {total} blocks, voice={voice}")
     print(f"{'='*60}")
 
+    from media_storage import save_audio
+
     for idx, block in enumerate(blocks):
         block_id = block['id']
         block_type = block['block_type'] if 'block_type' in block.keys() else 'paragraph'
@@ -437,11 +420,14 @@ def generate_chapter_audio():
             if source_format != 'mp3':
                 audio_base64 = convert_to_mp3(audio_base64, source_format)
 
+            # v4.3: ফাইলে সেভ
+            rel_path = save_audio(project_id, block_id, audio_base64, 'mp3') if project_id else None
+
             cursor.execute('''
                 UPDATE markdown_blocks 
-                SET audio_data = ?, audio_format = 'mp3', transcription = ?
+                SET audio_path = ?, audio_data = '', audio_format = 'mp3', transcription = ?
                 WHERE id = ?
-            ''', (audio_base64, json.dumps(transcription), block_id))
+            ''', (rel_path, json.dumps(transcription), block_id))
 
             results.append({
                 'block_id': block_id,

routes/project_routes.py

@@ -1,12 +1,18 @@
-# routes/project_routes.py - Project Management Routes (v4.2)
+# routes/project_routes.py - Project Management Routes (v4.3)
 
 import re
+import os
 import json
 import base64
-from flask import Blueprint, request, jsonify
+from flask import Blueprint, request, jsonify, send_file
 
-from db import get_db, vacuum_db
+from db import get_db, vacuum_db, get_db_stats
 from auth import login_required
+from media_storage import (
+    save_audio, save_image, save_thumbnail,
+    read_file_base64, get_safe_abs_path,
+    delete_project_media, get_storage_usage_bytes
+)
 
 project_bp = Blueprint('project', __name__)
 
@@ -48,6 +54,7 @@ def list_projects():
     cursor.execute('''
         SELECT p.id, p.name, p.created_at, p.updated_at,
                p.is_published, p.published_at, p.thumbnail_data, p.thumbnail_format,
+               p.thumbnail_path,
                p.description, p.author, p.category, p.view_count,
                (SELECT COUNT(*) FROM chapters WHERE project_id = p.id) as chapter_count,
                (SELECT COUNT(*) FROM markdown_blocks mb 
@@ -55,13 +62,20 @@ def list_projects():
                 WHERE c.project_id = p.id) as block_count,
                (SELECT COUNT(*) FROM markdown_blocks mb 
                 JOIN chapters c ON mb.chapter_id = c.id 
-                WHERE c.project_id = p.id AND mb.audio_data IS NOT NULL AND mb.audio_data != '') as audio_count
+                WHERE c.project_id = p.id 
+                AND ((mb.audio_data IS NOT NULL AND mb.audio_data != '')
+                     OR (mb.audio_path IS NOT NULL AND mb.audio_path != ''))) as audio_count
         FROM projects p
         ORDER BY p.updated_at DESC
     ''')
     
     projects = []
     for row in cursor.fetchall():
+        # thumbnail: path থাকলে ফাইল থেকে, নইলে পুরোনো base64
+        thumb_data = row['thumbnail_data']
+        if row['thumbnail_path']:
+            thumb_data = read_file_base64(row['thumbnail_path'])
+        
         projects.append({
             'id': row['id'],
             'name': row['name'],
@@ -72,7 +86,7 @@ def list_projects():
             'audio_count': row['audio_count'],
             'is_published': bool(row['is_published']),
             'published_at': row['published_at'],
-            'thumbnail_data': row['thumbnail_data'],
+            'thumbnail_data': thumb_data,
             'thumbnail_format': row['thumbnail_format'] or 'png',
             'description': row['description'] or '',
             'author': row['author'] or '',
@@ -115,9 +129,8 @@ def create_project():
 @login_required
 def get_project(project_id):
     """
-    Get project metadata WITHOUT audio_data.
-    Audio is loaded lazily via /api/projects/<id>/audio/<block_id>.
-    This keeps the response small (<1 MB) and avoids proxy truncation issues.
+    Get project metadata WITHOUT audio_data (lazy-loaded separately).
+    Images served as base64 from files (editor compatibility).
     """
     db = get_db()
     cursor = db.cursor()
@@ -137,8 +150,9 @@ def get_project(project_id):
     for chapter in chapters:
         cursor.execute('''
             SELECT id, block_order, block_type, content, tts_text,
-                   audio_format, transcription,
-                   (audio_data IS NOT NULL AND audio_data != '') as has_audio
+                   audio_format, audio_path, transcription,
+                   ((audio_data IS NOT NULL AND audio_data != '') 
+                    OR (audio_path IS NOT NULL AND audio_path != '')) as has_audio
             FROM markdown_blocks WHERE chapter_id = ? ORDER BY block_order
         ''', (chapter['id'],))
         blocks = cursor.fetchall()
@@ -146,7 +160,8 @@ def get_project(project_id):
         blocks_data = []
         for block in blocks:
             cursor.execute('''
-                SELECT * FROM block_images WHERE block_id = ? ORDER BY id
+                SELECT id, image_data, image_format, alt_text, position, image_path
+                FROM block_images WHERE block_id = ? ORDER BY id
             ''', (block['id'],))
             images = cursor.fetchall()
             
@@ -158,23 +173,33 @@ def get_project(project_id):
                 except (json.JSONDecodeError, TypeError):
                     transcription = []
             
+            images_data = []
+            for img in images:
+                # path থাকলে ফাইল থেকে, নইলে পুরোনো base64
+                img_data = ''
+                if img['image_path']:
+                    img_data = read_file_base64(img['image_path'])
+                elif img['image_data']:
+                    img_data = clean_str(img['image_data'])
+                images_data.append({
+                    'id': img['id'],
+                    'data': img_data,
+                    'format': clean_str(img['image_format']) or 'png',
+                    'alt_text': clean_str(img['alt_text']),
+                    'position': clean_str(img['position']) or 'before'
+                })
+            
             blocks_data.append({
                 'id': block['id'],
                 'block_order': block['block_order'],
                 'block_type': clean_str(block['block_type']),
                 'content': clean_str(block['content']),
                 'tts_text': clean_str(block['tts_text']),
-                'audio_data': '',  # Empty here; loaded lazily by frontend
+                'audio_data': '',
                 'audio_format': clean_str(block['audio_format']) or 'mp3',
                 'has_audio': bool(block['has_audio']),
                 'transcription': transcription,
-                'images': [{
-                    'id': img['id'],
-                    'data': clean_str(img['image_data']),
-                    'format': clean_str(img['image_format']) or 'png',
-                    'alt_text': clean_str(img['alt_text']),
-                    'position': clean_str(img['position']) or 'before'
-                } for img in images]
+                'images': images_data
             })
         
         chapters_data.append({
@@ -197,15 +222,12 @@ def get_project(project_id):
 @project_bp.route('/api/projects/<int:project_id>/audio/<int:block_id>', methods=['GET'])
 @login_required
 def get_block_audio(project_id, block_id):
-    """
-    Return audio_data (base64) for a single block.
-    Used by the frontend to lazy-load audio after metadata is loaded.
-    """
+    """Stream audio for a single block (v4.3: from file, with base64 fallback)."""
     db = get_db()
     cursor = db.cursor()
     
     cursor.execute('''
-        SELECT mb.audio_data, mb.audio_format
+        SELECT mb.audio_data, mb.audio_path, mb.audio_format
         FROM markdown_blocks mb
         JOIN chapters c ON mb.chapter_id = c.id
         WHERE mb.id = ? AND c.project_id = ?
@@ -215,13 +237,21 @@ def get_block_audio(project_id, block_id):
     if not row:
         return jsonify({'error': 'Block not found'}), 404
     
-    if not row['audio_data']:
-        return jsonify({'audio_data': '', 'audio_format': row['audio_format'] or 'mp3'})
+    # নতুন: ফাইল থেকে সরাসরি stream (Range request সাপোর্ট সহ)
+    if row['audio_path']:
+        abs_path = get_safe_abs_path(row['audio_path'])
+        if abs_path and os.path.exists(abs_path):
+            return send_file(abs_path, mimetype=f"audio/{row['audio_format'] or 'mp3'}",
+                             conditional=True)
     
-    return jsonify({
-        'audio_data': clean_str(row['audio_data']),
-        'audio_format': clean_str(row['audio_format']) or 'mp3'
-    })
+    # পুরোনো: base64 JSON
+    if row['audio_data']:
+        return jsonify({
+            'audio_data': clean_str(row['audio_data']),
+            'audio_format': clean_str(row['audio_format']) or 'mp3'
+        })
+    
+    return jsonify({'audio_data': '', 'audio_format': row['audio_format'] or 'mp3'})
 
 
 @project_bp.route('/api/projects/<int:project_id>', methods=['PUT'])
@@ -256,7 +286,7 @@ def update_project(project_id):
 @project_bp.route('/api/projects/<int:project_id>', methods=['DELETE'])
 @login_required
 def delete_project(project_id):
-    """Delete a project and all its data."""
+    """Delete a project, all DB data, AND its media folder (v4.3, no auto-vacuum)."""
     db = get_db()
     cursor = db.cursor()
     
@@ -282,7 +312,11 @@ def delete_project(project_id):
     cursor.execute('DELETE FROM projects WHERE id = ?', (project_id,))
     
     db.commit()
-    vacuum_db()
+    
+    # v4.3: প্রজেক্টের সব মিডিয়া ফাইল মুছি
+    delete_project_media(project_id)
+    
+    # NOTE: vacuum আর অটোমেটিক চলে না — ইউজার সেটিংস থেকে ম্যানুয়ালি করবে
     
     return jsonify({'success': True})
 
@@ -290,7 +324,7 @@ def delete_project(project_id):
 @project_bp.route('/api/projects/<int:project_id>/save', methods=['POST'])
 @login_required
 def save_project_content(project_id):
-    """Save all chapters and blocks for a project."""
+    """Save all chapters and blocks. Audio/images stored as FILES (v4.3)."""
     data = request.json
     chapters = data.get('chapters', [])
     
@@ -301,6 +335,7 @@ def save_project_content(project_id):
     if not cursor.fetchone():
         return jsonify({'error': 'Project not found'}), 404
     
+    # পুরোনো DB রেকর্ড মুছি (ফাইলগুলো নতুন করে লেখা হবে)
     cursor.execute('''
         DELETE FROM block_images WHERE block_id IN (
             SELECT mb.id FROM markdown_blocks mb
@@ -332,35 +367,58 @@ def save_project_content(project_id):
         
         for block in chapter.get('blocks', []):
             transcription = clean_transcription(block.get('transcription', []))
+            audio_format = clean_str(block.get('audio_format', 'mp3')) or 'mp3'
             
             cursor.execute('''
                 INSERT INTO markdown_blocks 
-                (chapter_id, block_order, block_type, content, tts_text, audio_data, audio_format, transcription)
-                VALUES (?, ?, ?, ?, ?, ?, ?, ?)
+                (chapter_id, block_order, block_type, content, tts_text, 
+                 audio_data, audio_path, audio_format, transcription)
+                VALUES (?, ?, ?, ?, ?, '', NULL, ?, ?)
             ''', (
                 chapter_id,
                 block['block_order'],
                 clean_str(block.get('block_type', 'paragraph')),
                 clean_str(block.get('content', '')),
                 clean_str(block.get('tts_text', '')),
-                clean_str(block.get('audio_data', '')),
-                clean_str(block.get('audio_format', 'mp3')),
+                audio_format,
                 json.dumps(transcription)
             ))
             
             block_id = cursor.lastrowid
             
+            # অডিও ফাইলে সেভ করি
+            audio_b64 = block.get('audio_data', '')
+            if audio_b64:
+                rel_path = save_audio(project_id, block_id, audio_b64, audio_format)
+                if rel_path:
+                    cursor.execute(
+                        'UPDATE markdown_blocks SET audio_path = ? WHERE id = ?',
+                        (rel_path, block_id)
+                    )
+            
+            # ইমেজগুলো ফাইলে সেভ করি
             for img in block.get('images', []):
+                img_format = clean_str(img.get('format', 'png')) or 'png'
                 cursor.execute('''
-                    INSERT INTO block_images (block_id, image_data, image_format, alt_text, position)
-                    VALUES (?, ?, ?, ?, ?)
+                    INSERT INTO block_images 
+                    (block_id, image_data, image_path, image_format, alt_text, position)
+                    VALUES (?, '', NULL, ?, ?, ?)
                 ''', (
                     block_id,
-                    clean_str(img.get('data', '')),
-                    clean_str(img.get('format', 'png')),
+                    img_format,
                     clean_str(img.get('alt_text', '')),
                     clean_str(img.get('position', 'before'))
                 ))
+                image_id = cursor.lastrowid
+                
+                img_b64 = img.get('data', '')
+                if img_b64:
+                    img_rel = save_image(project_id, image_id, img_b64, img_format)
+                    if img_rel:
+                        cursor.execute(
+                            'UPDATE block_images SET image_path = ? WHERE id = ?',
+                            (img_rel, image_id)
+                        )
     
     cursor.execute('''
         UPDATE projects SET updated_at = CURRENT_TIMESTAMP WHERE id = ?
@@ -392,7 +450,9 @@ def publish_project(project_id):
     cursor.execute('''
         SELECT COUNT(*) as cnt FROM markdown_blocks mb
         JOIN chapters c ON mb.chapter_id = c.id
-        WHERE c.project_id = ? AND mb.audio_data IS NOT NULL AND mb.audio_data != ''
+        WHERE c.project_id = ? 
+        AND ((mb.audio_data IS NOT NULL AND mb.audio_data != '')
+             OR (mb.audio_path IS NOT NULL AND mb.audio_path != ''))
     ''', (project_id,))
     audio_count = cursor.fetchone()['cnt']
     
@@ -440,7 +500,7 @@ def unpublish_project(project_id):
 @project_bp.route('/api/projects/<int:project_id>/thumbnail', methods=['POST'])
 @login_required
 def upload_thumbnail(project_id):
-    """Upload a thumbnail image."""
+    """Upload a thumbnail image (v4.3: stored as file)."""
     if 'file' not in request.files:
         return jsonify({'error': 'No file provided'}), 400
     
@@ -460,19 +520,21 @@ def upload_thumbnail(project_id):
     if fmt == 'jpg':
         fmt = 'jpeg'
     
-    b64 = base64.b64encode(img_bytes).decode('utf-8')
-    
     db = get_db()
     cursor = db.cursor()
     cursor.execute('SELECT id FROM projects WHERE id = ?', (project_id,))
     if not cursor.fetchone():
         return jsonify({'error': 'Project not found'}), 404
     
+    rel_path = save_thumbnail(project_id, img_bytes, fmt)
+    
     cursor.execute('''
-        UPDATE projects SET thumbnail_data = ?, thumbnail_format = ? WHERE id = ?
-    ''', (b64, fmt, project_id))
+        UPDATE projects SET thumbnail_path = ?, thumbnail_data = NULL, thumbnail_format = ? 
+        WHERE id = ?
+    ''', (rel_path, fmt, project_id))
     db.commit()
     
+    b64 = read_file_base64(rel_path)
     return jsonify({
         'success': True,
         'thumbnail_data': b64,
@@ -483,9 +545,51 @@ def upload_thumbnail(project_id):
 @project_bp.route('/api/projects/<int:project_id>/thumbnail', methods=['DELETE'])
 @login_required
 def delete_thumbnail(project_id):
-    """Remove project thumbnail."""
+    """Remove project thumbnail (DB + file)."""
     db = get_db()
     cursor = db.cursor()
-    cursor.execute('UPDATE projects SET thumbnail_data = NULL WHERE id = ?', (project_id,))
+    cursor.execute('SELECT thumbnail_path FROM projects WHERE id = ?', (project_id,))
+    row = cursor.fetchone()
+    if row and row['thumbnail_path']:
+        from media_storage import delete_file
+        delete_file(row['thumbnail_path'])
+    cursor.execute('UPDATE projects SET thumbnail_data = NULL, thumbnail_path = NULL WHERE id = ?',
+                   (project_id,))
     db.commit()
     return jsonify({'success': True})
+
+
+# ============================================
+# v4.3: Database Maintenance (VACUUM + stats)
+# ============================================
+
+@project_bp.route('/api/maintenance/db-stats', methods=['GET'])
+@login_required
+def db_stats():
+    """ডেটাবেস সাইজ, ফাঁকা স্পেস (%), এবং মিডিয়া স্টোরেজ সাইজ রিটার্ন করে।"""
+    stats = get_db_stats()
+    media_bytes = get_storage_usage_bytes()
+    stats['media_size_bytes'] = media_bytes
+    stats['media_size_mb'] = round(media_bytes / (1024 * 1024), 2)
+    return jsonify(stats)
+
+
+@project_bp.route('/api/maintenance/vacuum', methods=['POST'])
+@login_required
+def run_vacuum():
+    """ম্যানুয়ালি ডেটাবেস VACUUM চালায় (ফাঁকা স্পেস reclaim করে)।"""
+    before = get_db_stats()
+    try:
+        vacuum_db()
+    except Exception as e:
+        return jsonify({'error': f'VACUUM failed: {str(e)}'}), 500
+    after = get_db_stats()
+    
+    reclaimed_mb = round(before['file_size_mb'] - after['file_size_mb'], 2)
+    return jsonify({
+        'success': True,
+        'message': f'VACUUM complete. Reclaimed {reclaimed_mb} MB.',
+        'before': before,
+        'after': after,
+        'reclaimed_mb': reclaimed_mb
+    })

routes/public_routes.py

@@ -1,10 +1,12 @@
-# routes/public_routes.py - Public (No Auth) Routes for Published Audiobooks
+# routes/public_routes.py - Public (No Auth) Routes for Published Audiobooks (v4.3)
 
 import re
+import os
 import json
-from flask import Blueprint, jsonify, send_from_directory, abort
+from flask import Blueprint, jsonify, send_from_directory, send_file, abort
 
 from db import get_db
+from media_storage import get_safe_abs_path, read_file_base64
 
 public_bp = Blueprint('public', __name__)
 
@@ -53,7 +55,6 @@ def public_reader(project_id):
     if not project or not project['is_published']:
         abort(404)
     
-    # Increment view count
     cursor.execute('UPDATE projects SET view_count = view_count + 1 WHERE id = ?', (project_id,))
     db.commit()
     
@@ -68,7 +69,7 @@ def list_published_books():
     
     cursor.execute('''
         SELECT p.id, p.name, p.description, p.author, p.category,
-               p.thumbnail_data, p.thumbnail_format, p.published_at,
+               p.thumbnail_data, p.thumbnail_format, p.thumbnail_path, p.published_at,
                p.view_count, p.created_at,
                (SELECT COUNT(*) FROM chapters WHERE project_id = p.id) as chapter_count
         FROM projects p
@@ -78,13 +79,17 @@ def list_published_books():
     
     books = []
     for row in cursor.fetchall():
+        thumb_data = row['thumbnail_data']
+        if row['thumbnail_path']:
+            thumb_data = read_file_base64(row['thumbnail_path'])
+        
         books.append({
             'id': row['id'],
             'name': row['name'],
             'description': row['description'] or '',
             'author': row['author'] or '',
             'category': row['category'] or '',
-            'thumbnail_data': row['thumbnail_data'],
+            'thumbnail_data': thumb_data,
             'thumbnail_format': row['thumbnail_format'] or 'png',
             'published_at': row['published_at'],
             'view_count': row['view_count'] or 0,
@@ -96,11 +101,7 @@ def list_published_books():
 
 @public_bp.route('/api/public/books/<int:project_id>', methods=['GET'])
 def get_published_book(project_id):
-    """
-    Get book metadata WITHOUT audio_data.
-    Audio is loaded lazily via /api/public/books/<id>/audio/<block_id>.
-    This keeps the response small (<1 MB) and avoids proxy truncation issues.
-    """
+    """Get book metadata WITHOUT audio_data (lazy-loaded separately)."""
     db = get_db()
     cursor = db.cursor()
     
@@ -120,8 +121,9 @@ def get_published_book(project_id):
     chapters_data = []
     for chapter in chapters:
         cursor.execute('''
-            SELECT id, block_order, block_type, content, audio_format, transcription,
-                   (audio_data IS NOT NULL AND audio_data != '') as has_audio
+            SELECT id, block_order, block_type, content, audio_format, audio_path, transcription,
+                   ((audio_data IS NOT NULL AND audio_data != '')
+                    OR (audio_path IS NOT NULL AND audio_path != '')) as has_audio
             FROM markdown_blocks WHERE chapter_id = ? ORDER BY block_order
         ''', (chapter['id'],))
         blocks = cursor.fetchall()
@@ -129,7 +131,8 @@ def get_published_book(project_id):
         blocks_data = []
         for block in blocks:
             cursor.execute('''
-                SELECT * FROM block_images WHERE block_id = ? ORDER BY id
+                SELECT image_data, image_format, alt_text, position, image_path
+                FROM block_images WHERE block_id = ? ORDER BY id
             ''', (block['id'],))
             images = cursor.fetchall()
             
@@ -141,21 +144,30 @@ def get_published_book(project_id):
                 except (json.JSONDecodeError, TypeError):
                     transcription = []
             
+            images_data = []
+            for img in images:
+                img_data = ''
+                if img['image_path']:
+                    img_data = read_file_base64(img['image_path'])
+                elif img['image_data']:
+                    img_data = clean_str(img['image_data'])
+                images_data.append({
+                    'data': img_data,
+                    'format': clean_str(img['image_format']) or 'png',
+                    'alt_text': clean_str(img['alt_text']),
+                    'position': clean_str(img['position']) or 'before'
+                })
+            
             blocks_data.append({
                 'id': block['id'],
                 'block_order': block['block_order'],
                 'block_type': clean_str(block['block_type']),
                 'content': clean_str(block['content']),
-                'audio_data': '',  # Empty here; loaded lazily by frontend
+                'audio_data': '',
                 'audio_format': clean_str(block['audio_format']) or 'mp3',
                 'has_audio': bool(block['has_audio']),
                 'transcription': transcription,
-                'images': [{
-                    'data': clean_str(img['image_data']),
-                    'format': clean_str(img['image_format']) or 'png',
-                    'alt_text': clean_str(img['alt_text']),
-                    'position': clean_str(img['position']) or 'before'
-                } for img in images]
+                'images': images_data
             })
         
         chapters_data.append({
@@ -165,12 +177,16 @@ def get_published_book(project_id):
             'blocks': blocks_data
         })
     
+    thumb_data = project['thumbnail_data']
+    if project['thumbnail_path']:
+        thumb_data = read_file_base64(project['thumbnail_path'])
+    
     return jsonify({
         'id': project['id'],
         'name': clean_str(project['name']),
         'description': clean_str(project['description']) if project['description'] else '',
         'author': clean_str(project['author']) if project['author'] else '',
-        'thumbnail_data': project['thumbnail_data'],
+        'thumbnail_data': thumb_data,
         'thumbnail_format': project['thumbnail_format'] or 'png',
         'chapters': chapters_data
     })
@@ -178,21 +194,17 @@ def get_published_book(project_id):
 
 @public_bp.route('/api/public/books/<int:project_id>/audio/<int:block_id>', methods=['GET'])
 def get_public_block_audio(project_id, block_id):
-    """
-    Return audio_data (base64) for a single block in a published book.
-    No auth required since the book is published publicly.
-    """
+    """Stream audio file for a published book block (v4.3)."""
     db = get_db()
     cursor = db.cursor()
     
-    # Verify project is published
     cursor.execute('SELECT is_published FROM projects WHERE id = ?', (project_id,))
     project = cursor.fetchone()
     if not project or not project['is_published']:
         return jsonify({'error': 'Book not found or not published'}), 404
     
     cursor.execute('''
-        SELECT mb.audio_data, mb.audio_format
+        SELECT mb.audio_data, mb.audio_path, mb.audio_format
         FROM markdown_blocks mb
         JOIN chapters c ON mb.chapter_id = c.id
         WHERE mb.id = ? AND c.project_id = ?
@@ -202,10 +214,16 @@ def get_public_block_audio(project_id, block_id):
     if not row:
         return jsonify({'error': 'Block not found'}), 404
     
-    if not row['audio_data']:
-        return jsonify({'audio_data': '', 'audio_format': row['audio_format'] or 'mp3'})
+    if row['audio_path']:
+        abs_path = get_safe_abs_path(row['audio_path'])
+        if abs_path and os.path.exists(abs_path):
+            return send_file(abs_path, mimetype=f"audio/{row['audio_format'] or 'mp3'}",
+                             conditional=True)
     
-    return jsonify({
-        'audio_data': clean_str(row['audio_data']),
-        'audio_format': clean_str(row['audio_format']) or 'mp3'
-    })
+    if row['audio_data']:
+        return jsonify({
+            'audio_data': clean_str(row['audio_data']),
+            'audio_format': clean_str(row['audio_format']) or 'mp3'
+        })
+    
+    return jsonify({'audio_data': '', 'audio_format': row['audio_format'] or 'mp3'})

static/js/app.js

@@ -1,6 +1,6 @@
 /**
- * Audiobook Maker Pro v4.2 - Main Application
- * UPDATED: Lazy audio loading to avoid large response truncation
+ * Audiobook Maker Pro v4.3 - Main Application
+ * UPDATED: Lazy audio loading + Storage & Maintenance (VACUUM)
  */
 
 // ============================================
@@ -17,6 +17,7 @@ let voices = [];
 let archiveModal = null;
 let ttsEditModal = null;
 let publishModal = null;
+let dbMaintenanceModal = null;
 let publishingProjectId = null;
 let currentWorkflowStage = 'upload';
 let allArchiveProjects = [];
@@ -26,7 +27,7 @@ let allArchiveProjects = [];
 // ============================================
 
 document.addEventListener('DOMContentLoaded', function() {
-    console.log('🎧 Audiobook Maker Pro v4.2 initializing...');
+    console.log('🎧 Audiobook Maker Pro v4.3 initializing...');
     
     archiveModal = new bootstrap.Modal(document.getElementById('archiveModal'));
     ttsEditModal = new bootstrap.Modal(document.getElementById('ttsEditModal'));
@@ -979,6 +980,32 @@ async function loadAudioBlocksInBackground(projectId, blockIds) {
     async function fetchOne(blockId) {
         try {
             const resp = await fetch(`/api/projects/${projectId}/audio/${blockId}`);
+            
+            // v4.3: এন্ডপয়েন্ট বাইনারি অডিও (audio/*) অথবা legacy base64 JSON দিতে পারে
+            const contentType = resp.headers.get('content-type') || '';
+            
+            const blockData = editorBlocks.find(b => b.db_id === blockId);
+            
+            if (contentType.startsWith('audio/')) {
+                // নতুন: ফাইল আছে — শুধু indicator আপডেট করি, base64 মেমরিতে রাখি না
+                // (reader নিজেই lazy fetch করবে)
+                if (blockData) {
+                    blockData.has_audio = true;
+                    const blockEl = document.getElementById(blockData.id);
+                    if (blockEl) {
+                        const indicator = blockEl.querySelector('.audio-indicator');
+                        if (indicator) {
+                            indicator.classList.remove('no-audio');
+                            indicator.classList.add('has-audio');
+                            indicator.title = 'Audio available';
+                        }
+                    }
+                }
+                loaded++;
+                return;
+            }
+            
+            // Legacy base64 JSON
             const data = await resp.json();
             
             if (data.error || !data.audio_data) {
@@ -986,13 +1013,11 @@ async function loadAudioBlocksInBackground(projectId, blockIds) {
                 return;
             }
             
-            // Update editorBlocks state by db_id
-            const blockData = editorBlocks.find(b => b.db_id === blockId);
             if (blockData) {
                 blockData.audio_data = data.audio_data;
                 blockData.audio_format = data.audio_format;
+                blockData.has_audio = true;
                 
-                // Update DOM indicator (green dot)
                 const blockEl = document.getElementById(blockData.id);
                 if (blockEl) {
                     const indicator = blockEl.querySelector('.audio-indicator');
@@ -1017,8 +1042,8 @@ async function loadAudioBlocksInBackground(projectId, blockIds) {
     }
     
     const msg = failed > 0 
-        ? `Loaded ${loaded}/${blockIds.length} audio blocks (${failed} failed)`
-        : `All ${loaded} audio blocks loaded ✓`;
+        ? `Verified ${loaded}/${blockIds.length} audio blocks (${failed} failed)`
+        : `All ${loaded} audio blocks verified ✓`;
     showNotification(msg, failed > 0 ? 'warning' : 'success');
     
     if (typeof updatePanelUI === 'function') {
@@ -1028,7 +1053,7 @@ async function loadAudioBlocksInBackground(projectId, blockIds) {
 
 
 async function deleteProject(projectId) {
-    if (!confirm('Are you sure you want to delete this project? This action cannot be undone.')) return;
+    if (!confirm('Are you sure you want to delete this project? This action cannot be undone.\n\nএই প্রজেক্টের অডিও ও ইমেজ ফাইলগুলোও মুছে যাবে।')) return;
     
     showLoader('Deleting...');
     
@@ -1046,6 +1071,111 @@ async function deleteProject(projectId) {
     }
 }
 
+// ============================================
+// v4.3: Storage & Maintenance (VACUUM)
+// ============================================
+
+function openDbMaintenance() {
+    if (!dbMaintenanceModal) {
+        dbMaintenanceModal = new bootstrap.Modal(document.getElementById('dbMaintenanceModal'));
+    }
+    dbMaintenanceModal.show();
+    loadDbStats();
+}
+
+async function loadDbStats() {
+    const loadingEl = document.getElementById('dbStatsLoading');
+    const contentEl = document.getElementById('dbStatsContent');
+    if (loadingEl) loadingEl.style.display = 'block';
+    if (contentEl) contentEl.style.display = 'none';
+    
+    try {
+        const resp = await fetch('/api/maintenance/db-stats');
+        const s = await resp.json();
+        
+        if (s.error) {
+            showNotification(s.error, 'error');
+            return;
+        }
+        
+        document.getElementById('dbmFileSize').textContent = `${s.file_size_mb} MB`;
+        document.getElementById('dbmMediaSize').textContent = `${s.media_size_mb} MB`;
+        document.getElementById('dbmFreeText').textContent =
+            `${s.free_mb} MB (${s.free_percent}%)`;
+        
+        const bar = document.getElementById('dbmFreeBar');
+        const pct = Math.min(s.free_percent, 100);
+        bar.style.width = pct + '%';
+        bar.textContent = `${s.free_percent}%`;
+        
+        // রঙ: কম হলে সবুজ, বেশি হলে হলুদ/লাল
+        bar.className = 'progress-bar';
+        if (s.free_percent >= 30) {
+            bar.classList.add('bg-danger');
+        } else if (s.free_percent >= 15) {
+            bar.classList.add('bg-warning');
+        } else {
+            bar.classList.add('bg-success');
+        }
+        
+        const advice = document.getElementById('dbmAdvice');
+        if (s.free_percent >= 15) {
+            advice.className = 'alert alert-warning';
+            advice.style.display = 'block';
+            advice.innerHTML = `<i class="bi bi-exclamation-triangle me-1"></i>` +
+                `ডেটাবেসে <strong>${s.free_percent}%</strong> ফাঁকা স্পেস জমেছে। ` +
+                `<strong>Run VACUUM</strong> চালিয়ে এটি reclaim করতে পারেন।`;
+        } else {
+            advice.className = 'alert alert-success';
+            advice.style.display = 'block';
+            advice.innerHTML = `<i class="bi bi-check-circle me-1"></i>` +
+                `ফাঁকা স্পেস কম (<strong>${s.free_percent}%</strong>) — এখন VACUUM চালানোর দরকার নেই।`;
+        }
+        
+        if (loadingEl) loadingEl.style.display = 'none';
+        if (contentEl) contentEl.style.display = 'block';
+        
+    } catch (e) {
+        console.error(e);
+        showNotification('Failed to load storage info', 'error');
+    }
+}
+
+async function runDbVacuum() {
+    const vacuumBtn = document.getElementById('dbmVacuumBtn');
+    const refreshBtn = document.getElementById('dbmRefreshBtn');
+    
+    if (!confirm('VACUUM এখন চালাবেন? এটি ডেটাবেস ছোট করবে কিন্তু কিছু সময় (ডেটাবেস বড় হলে কয়েক মিনিট) নিতে পারে।')) {
+        return;
+    }
+    
+    if (vacuumBtn) {
+        vacuumBtn.disabled = true;
+        vacuumBtn.innerHTML = '<span class="spinner-border spinner-border-sm me-1"></span>Running...';
+    }
+    if (refreshBtn) refreshBtn.disabled = true;
+    
+    try {
+        const resp = await fetch('/api/maintenance/vacuum', { method: 'POST' });
+        const data = await resp.json();
+        
+        if (data.error) {
+            showNotification(data.error, 'error');
+        } else {
+            showNotification(data.message || 'VACUUM complete', 'success');
+            loadDbStats();
+        }
+    } catch (e) {
+        showNotification('VACUUM failed', 'error');
+    } finally {
+        if (vacuumBtn) {
+            vacuumBtn.disabled = false;
+            vacuumBtn.innerHTML = '<i class="bi bi-stars me-1"></i>Run VACUUM';
+        }
+        if (refreshBtn) refreshBtn.disabled = false;
+    }
+}
+
 // ============================================
 // TTS Text Editing
 // ============================================

static/js/interactive-reader.js

@@ -1,12 +1,12 @@
 /**
- * Interactive Reader Module — Lazy Audio Loading (v4)
+ * Interactive Reader Module — File-based Audio (v4.3)
  *
  * Strategy:
- *   - Text + transcription are already loaded (from editorBlocks in memory).
- *   - Audio is fetched on-demand from /api/projects/<id>/audio/<block_id>
- *     when the user wants to play that block.
- *   - Smart preload: at 70% of current block, fetch next block's audio.
- *   - Memory cap: keep at most MAX_AUDIO_LOADED blob URLs alive.
+ *   - Text + transcription loaded from editorBlocks in memory.
+ *   - Audio fetched on-demand from /api/projects/<id>/audio/<block_id>.
+ *     v4.3: endpoint may return binary audio (Content-Type: audio/*) OR
+ *     legacy base64 JSON. Both handled.
+ *   - Smart preload + memory cap (sliding window of blob URLs).
  */
 
 // ============================================
@@ -66,7 +66,6 @@ function renderInteractiveReader() {
 
             isFirstBlockOfChapter = false;
 
-            // has_audio comes from server; audio_data may not yet be loaded
             if (!isImageBlock && blockData && (blockData.audio_data || blockData.has_audio)) {
                 hasAudio = true;
             }
@@ -106,7 +105,6 @@ function renderInteractiveReader() {
         const blockData = block._editorData;
         const isImageBlock = block._isImage;
 
-        // has_audio is the SOURCE OF TRUTH for whether this block has audio on server
         const hasBlockAudio = !isImageBlock && blockData && (blockData.audio_data || blockData.has_audio);
         const blockId = blockData ? blockData.id : `reader_${Date.now()}_${Math.random().toString(36).substr(2, 5)}`;
 
@@ -447,15 +445,16 @@ function setReaderButtonLoading(isLoading) {
 }
 
 // ============================================
-// Audio Lazy Loading
+// Audio Lazy Loading (v4.3 — binary OR base64)
 // ============================================
 
 /**
- * Fetch audio for an instance. If already loaded into editorBlocks
- * by background loader, use that. Otherwise fetch from API directly.
+ * Fetch audio for an instance.
+ * v4.3: endpoint may return binary audio (Content-Type: audio/*)
+ * → return { audio_url } ; OR legacy base64 JSON → return { audio_data }.
  */
 async function fetchAudioForInstance(inst) {
-    // Path 1: audio_data already in editorBlocks (loaded in background)
+    // Path 1: audio_data already in editorBlocks (rare, generated this session)
     if (inst.blockData && inst.blockData.audio_data) {
         return {
             audio_data: inst.blockData.audio_data,
@@ -469,16 +468,22 @@ async function fetchAudioForInstance(inst) {
     }
 
     const resp = await fetch(`/api/projects/${currentProject.id}/audio/${inst.blockData.db_id}`);
-    const data = await resp.json();
+    if (!resp.ok) throw new Error('No audio data');
 
+    const contentType = resp.headers.get('content-type') || '';
+    if (contentType.startsWith('audio/')) {
+        // v4.3: direct binary stream
+        const blob = await resp.blob();
+        return { audio_url: URL.createObjectURL(blob), audio_format: 'mp3' };
+    }
+
+    // Legacy base64 JSON
+    const data = await resp.json();
     if (data.error || !data.audio_data) {
         throw new Error(data.error || 'No audio data');
     }
-    
-    // Cache into editorBlocks for future use
     inst.blockData.audio_data = data.audio_data;
     inst.blockData.audio_format = data.audio_format;
-    
     return data;
 }
 
@@ -488,8 +493,13 @@ function ensureReaderAudioLoaded(inst) {
 
     inst.audioLoadingPromise = (async () => {
         const audioInfo = await fetchAudioForInstance(inst);
-        const audioBlob = base64ToBlob(audioInfo.audio_data, `audio/${audioInfo.audio_format || 'mp3'}`);
-        const audioUrl = URL.createObjectURL(audioBlob);
+        let audioUrl;
+        if (audioInfo.audio_url) {
+            audioUrl = audioInfo.audio_url; // v4.3 binary blob URL
+        } else {
+            const audioBlob = base64ToBlob(audioInfo.audio_data, `audio/${audioInfo.audio_format || 'mp3'}`);
+            audioUrl = URL.createObjectURL(audioBlob);
+        }
         const audio = new Audio(audioUrl);
 
         return new Promise((resolve, reject) => {

templates/index.html

@@ -3,7 +3,7 @@
 <head>
     <meta charset="UTF-8">
     <meta name="viewport" content="width=device-width, initial-scale=1.0">
-    <title>Audiobook Maker Pro v4.1</title>
+    <title>Audiobook Maker Pro v4.3</title>
     
     <link href="https://cdn.jsdelivr.net/npm/bootstrap@5.3.2/dist/css/bootstrap.min.css" rel="stylesheet">
     <link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/bootstrap-icons@1.11.1/font/bootstrap-icons.css">
@@ -163,7 +163,7 @@
                 <h1 class="app-title">
                     <i class="bi bi-soundwave"></i>
                     Audiobook Maker Pro
-                    <span class="version-badge">v4.2</span>
+                    <span class="version-badge">v4.3</span>
                 </h1>
             </div>
             <div class="header-right">
@@ -197,6 +197,11 @@
                                 </a>
                             </li>
                             <li id="adminDivider" style="display:none;"><hr class="dropdown-divider"></li>
+                            <li>
+                                <a class="dropdown-item" href="#" onclick="openDbMaintenance()">
+                                    <i class="bi bi-hdd-stack me-2"></i>Storage & Maintenance
+                                </a>
+                            </li>
                             <li>
                                 <a class="dropdown-item" href="#" onclick="openChangePassword()">
                                     <i class="bi bi-key me-2"></i>Change Password
@@ -467,6 +472,69 @@
         </div>
     </div>
 
+    <!-- v4.3: Storage & Maintenance Modal -->
+    <div class="modal fade" id="dbMaintenanceModal" tabindex="-1">
+        <div class="modal-dialog">
+            <div class="modal-content">
+                <div class="modal-header">
+                    <h5 class="modal-title">
+                        <i class="bi bi-hdd-stack me-2"></i>Storage & Maintenance
+                    </h5>
+                    <button type="button" class="btn-close" data-bs-dismiss="modal"></button>
+                </div>
+                <div class="modal-body">
+                    <div id="dbStatsLoading" class="text-center py-4">
+                        <div class="spinner-border text-primary" role="status"></div>
+                        <p class="mt-2 text-muted">Loading storage info...</p>
+                    </div>
+
+                    <div id="dbStatsContent" style="display:none;">
+                        <div class="row g-3 mb-3">
+                            <div class="col-6">
+                                <div class="p-3 rounded" style="background: var(--bg-tertiary);">
+                                    <div class="text-muted small mb-1"><i class="bi bi-database me-1"></i>Database Size</div>
+                                    <div class="fs-5 fw-bold" id="dbmFileSize">— MB</div>
+                                </div>
+                            </div>
+                            <div class="col-6">
+                                <div class="p-3 rounded" style="background: var(--bg-tertiary);">
+                                    <div class="text-muted small mb-1"><i class="bi bi-folder me-1"></i>Media Files Size</div>
+                                    <div class="fs-5 fw-bold" id="dbmMediaSize">— MB</div>
+                                </div>
+                            </div>
+                        </div>
+
+                        <div class="mb-2 d-flex justify-content-between align-items-center">
+                            <span class="small fw-semibold">
+                                <i class="bi bi-trash3 me-1"></i>Reclaimable free space
+                            </span>
+                            <span class="small text-muted" id="dbmFreeText">— MB (—%)</span>
+                        </div>
+                        <div class="progress mb-3" style="height: 14px;">
+                            <div class="progress-bar" id="dbmFreeBar" role="progressbar" style="width: 0%;"></div>
+                        </div>
+
+                        <div class="alert" id="dbmAdvice" style="display:none;"></div>
+
+                        <p class="text-muted small mb-0">
+                            <i class="bi bi-info-circle me-1"></i>
+                            VACUUM ডেটাবেসের ফাঁকা স্পেস reclaim করে এটি ছোট করে। প্রজেক্ট ডিলিট করার পর বা মাসে একবার চালানো ভালো। এটি কিছু সময় নিতে পারে।
+                        </p>
+                    </div>
+                </div>
+                <div class="modal-footer">
+                    <button type="button" class="btn btn-secondary" data-bs-dismiss="modal">Close</button>
+                    <button type="button" class="btn btn-outline-secondary" id="dbmRefreshBtn" onclick="loadDbStats()">
+                        <i class="bi bi-arrow-clockwise me-1"></i>Refresh
+                    </button>
+                    <button type="button" class="btn btn-primary" id="dbmVacuumBtn" onclick="runDbVacuum()">
+                        <i class="bi bi-stars me-1"></i>Run VACUUM
+                    </button>
+                </div>
+            </div>
+        </div>
+    </div>
+
     <script src="https://cdn.jsdelivr.net/npm/bootstrap@5.3.2/dist/js/bootstrap.bundle.min.js"></script>
     <script src="https://cdn.jsdelivr.net/npm/marked/marked.min.js"></script>
     <script src="/static/js/app.js"></script>