tts2 / app.py
Sada8888's picture
Update app.py
ed9d19c verified
Raw
History Blame Contribute Delete
33.7 kB
import os
import json
import time
import io
import threading
import uuid
import requests
import re
import logging
import random
import base64
import sqlite3
import concurrent.futures
from datetime import datetime, timedelta
from itertools import cycle
from flask import Flask, request, jsonify, render_template, send_file
from flask_cors import CORS
from pydub import AudioSegment
# --- CONFIGURATION & LOGGING ---
CACHE_DIRECTORY = "/tmp/huggingface_cache_ezmary"
os.makedirs(CACHE_DIRECTORY, exist_ok=True)
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s', datefmt='%Y-%m-%d %H:%M:%S')
app = Flask(__name__)
CORS(app)
# --- WORKER POOL SETUP ---
WORKER_URLS = [
"https://opera8-ttspro.hf.space/generate",
"https://opera8-ttspro1.hf.space/generate",
"https://hamed744-ttspro.hf.space/generate",
"https://hamed744-ttspro2.hf.space/generate",
]
# --- آدرس سرویس تغییر صدا (VC) ---
VC_SPACE_URL = "https://ezmarynoori-sada.hf.space"
# --- GLOBAL VARIABLES ---
tasks = {}
tasks_lock = threading.Lock()
USAGE_LIMIT = 5
# --- BROWSER HEADERS (Anti-Bot Bypass) ---
STANDARD_HEADERS = {
"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36",
"Accept": "application/json",
"Content-Type": "application/json"
}
# --- DATABASE LOGIC (SQLite in Persistent Storage) ---
DB_DIR = "/data" if os.path.isdir("/data") else "."
DB_PATH = os.path.join(DB_DIR, "podcast.db")
def get_db_connection():
conn = sqlite3.connect(DB_PATH, timeout=20)
conn.row_factory = sqlite3.Row
return conn
def init_db():
try:
with get_db_connection() as conn:
conn.execute('''
CREATE TABLE IF NOT EXISTS users (
fingerprint TEXT PRIMARY KEY,
ips TEXT NOT NULL,
count INTEGER NOT NULL,
week_start REAL NOT NULL
)
''')
conn.commit()
logging.info(f"✅ دیتابیس SQLite با موفقیت در مسیر زیر متصل شد: {DB_PATH}")
except Exception as e:
logging.error(f"❌ خطا در ساخت دیتابیس SQLite: {e}")
def get_user_record(conn, fingerprint, ip):
cursor = conn.execute('SELECT * FROM users WHERE fingerprint = ?', (fingerprint,))
row = cursor.fetchone()
if row:
return dict(row)
ip_search = f'%"{ip}"%'
cursor = conn.execute('SELECT * FROM users WHERE ips LIKE ?', (ip_search,))
row = cursor.fetchone()
if row:
return dict(row)
return None
def background_cleanup():
while True:
time.sleep(86400)
try:
six_months_ago = time.time() - (6 * 30 * 24 * 60 * 60)
with get_db_connection() as conn:
conn.execute('DELETE FROM users WHERE week_start < ?', (six_months_ago,))
conn.commit()
except Exception as e:
logging.error(f"DB Cleanup error: {e}")
def get_user_ip():
if request.headers.getlist("X-Forwarded-For"):
return request.headers.getlist("X-Forwarded-For")[0].split(',')[0].strip()
return request.remote_addr
# --- TTS HELPER FUNCTIONS ---
def call_worker(index, chunk_payload):
raw_text = chunk_payload.get("text", "")
clean_text = re.sub(r'[*_`~#]', '', raw_text)
clean_text = re.sub(r'\[.*?\]|\(.*?\)', '', clean_text)
clean_text = clean_text.strip()
if not clean_text:
return index, AudioSegment.silent(duration=500)
use_live = True if len(clean_text) <= 600 else False
target_speaker = chunk_payload.get("speaker")
is_custom = chunk_payload.get("is_custom", False)
actual_speaker_request = "Charon" if is_custom else target_speaker
worker_payload = {
"text": clean_text,
"speaker": actual_speaker_request,
"temperature": chunk_payload.get("temperature", 0.9),
"use_live_model": use_live,
"retry_limit": 50,
"fallback_to_live": True
}
max_attempts = 40
for attempt in range(max_attempts):
workers = list(WORKER_URLS)
random.shuffle(workers)
for worker_url in workers:
try:
logging.info(f"▶️ [قطعه {index+1}] ارسال به {worker_url} (تلاش {attempt+1}/{max_attempts})")
response = requests.post(worker_url, json=worker_payload, headers=STANDARD_HEADERS, timeout=120)
if response.status_code == 200:
logging.info(f"✅ [قطعه {index+1}] با موفقیت ساخته شد.")
audio_data = io.BytesIO(response.content)
audio_segment = AudioSegment.from_file(audio_data)
return index, audio_segment
else:
logging.warning(f"⚠️ [قطعه {index+1}] ارور {response.status_code} از کارگر دریافت شد. تلاش بعدی...")
continue
except Exception as e:
logging.error(f"❌ [قطعه {index+1}] قطعی ارتباط: {str(e)[:40]}")
continue
sleep_time = 3 + (attempt * 1.5) + random.uniform(0.5, 2.0)
logging.warning(f"🔄 [قطعه {index+1}] کارگرها شلوغ/خطا دادند. استراحت به مدت {sleep_time:.1f} ثانیه...")
time.sleep(sleep_time)
raise ValueError(f"قطعه {index+1} پس از 40 بار تلاش ساخته نشد.")
# --- AI PODCAST SCRIPT LOGIC (Cohere Labs Space) ---
def generate_podcast_in_background(task_id, system_prompt, safety_settings):
try:
AYA_SPACE_URL = "https://coherelabs-aya-expanse.hf.space/gradio_api"
MAX_ATTEMPTS = 50
for attempt in range(MAX_ATTEMPTS):
try:
session_hash = str(uuid.uuid4())[0:11]
payload = {
"data": [system_prompt, [], None, None],
"event_data": None, "fn_index": 2, "session_hash": session_hash, "trigger_id": 37
}
join_response = requests.post(f"{AYA_SPACE_URL}/queue/join", json=payload, headers=STANDARD_HEADERS, timeout=60)
join_response.raise_for_status()
stream_url = f"{AYA_SPACE_URL}/queue/data?session_hash={session_hash}"
raw_text = None
with requests.get(stream_url, stream=True, headers=STANDARD_HEADERS, timeout=120) as resp:
for line in resp.iter_lines():
if line:
decoded_line = line.decode('utf-8')
if decoded_line.startswith('data: '):
json_str = decoded_line[6:]
try:
msg_data = json.loads(json_str)
msg_type = msg_data.get('msg')
if msg_type == 'process_completed':
output = msg_data.get('output', {}).get('data', [])
if output and len(output) > 0:
updated_history = output[0]
if updated_history:
raw_text = updated_history[-1][1]
elif msg_type == 'queue_full':
raise Exception("صف سرور شلوغ است")
except json.JSONDecodeError: continue
if not raw_text: raise ValueError("بدون پاسخ از مدل هوش مصنوعی")
json_string = None
json_pattern = r"`{3}json\s*(\{.*?\})\s*`{3}"
match = re.search(json_pattern, raw_text, re.DOTALL)
if match: json_string = match.group(1)
else:
s_idx = raw_text.find('{')
e_idx = raw_text.rfind('}')
if s_idx != -1 and e_idx != -1: json_string = raw_text[s_idx:e_idx+1]
if not json_string: raise ValueError("No JSON found in response")
data = json.loads(json_string)
if "script" in data:
for t in data["script"]:
if "dialogue" in t:
t["dialogue"] = re.sub(r'\[.*?\]|\(.*?\)', '', t["dialogue"]).strip()
with tasks_lock:
tasks[task_id].update({'status': 'completed', 'data': data})
return
except Exception as e:
logging.warning(f"AI Attempt {attempt} failed: {e}")
time.sleep(2)
with tasks_lock: tasks[task_id].update({'status': 'failed', 'error': 'Max retries reached'})
except Exception as e:
with tasks_lock: tasks[task_id].update({'status': 'failed', 'error': str(e)})
# --- FULL AUTO PODCAST LOGIC ---
def generate_full_podcast_audio_background(task_id, prompt, speakers):
try:
logging.info(f"🚀 [پروژه {task_id}] عملیات ساخت پادکست آغاز شد.")
with tasks_lock:
tasks[task_id] = {'status': 'writing_script', 'progress': 'در حال نگارش سناریو...'}
valid_speaker_ids = [str(s['id']).strip() for s in speakers]
default_speaker_id = valid_speaker_ids[0] if valid_speaker_ids else "Charon"
spk_text = "\n".join([f"- {s['id']}: {s['name']}" for s in speakers])
system_prompt = f"""Act as a Professional Podcast Producer.
Topic: "{prompt}"
Speakers Available:
{spk_text}
CRITICAL INSTRUCTION: You must create a VERY LONG, in-depth, and highly detailed podcast script.
- Do NOT write a short summary.
- The conversation must deeply explore the topic. Make it a very long and detailed conversation, as long as necessary to fully cover the topic.
- Make the dialogue engaging and informative. You must use two speakers for the podcast, or three if necessary, or whatever number of speakers the user specifies. The podcast should be an engaging exchange, typically between a man and a woman, with both speakers discussing the topic together.
Output ONLY valid JSON.
Format: {{"selected_speakers": ["id1", "id2"], "script": [{{"speaker_id": "id1", "dialogue": "..."}}]}}
Dialogue rules: Do NOT avoid writing stage directions and emotional cues like laughing or sighing. They are allowed, but the mood/action MUST be placed inside parentheses, for example (laugh) or (sigh)."""
AYA_SPACE_URL = "https://coherelabs-aya-expanse.hf.space/gradio_api"
MAX_ATTEMPTS = 50
data = None
logging.info("📝 در حال ارتباط با هوش مصنوعی برای نوشتن سناریو...")
for attempt in range(MAX_ATTEMPTS):
try:
session_hash = str(uuid.uuid4())[0:11]
payload = {"data": [system_prompt, [], None, None], "event_data": None, "fn_index": 2, "session_hash": session_hash, "trigger_id": 37}
join_response = requests.post(f"{AYA_SPACE_URL}/queue/join", json=payload, headers=STANDARD_HEADERS, timeout=60)
join_response.raise_for_status()
stream_url = f"{AYA_SPACE_URL}/queue/data?session_hash={session_hash}"
raw_text = None
with requests.get(stream_url, stream=True, headers=STANDARD_HEADERS, timeout=120) as resp:
for line in resp.iter_lines():
if line:
decoded_line = line.decode('utf-8')
if decoded_line.startswith('data: '):
json_str = decoded_line[6:]
try:
msg_data = json.loads(json_str)
if msg_data.get('msg') == 'process_completed':
output = msg_data.get('output', {}).get('data', [])
if output and len(output) > 0 and output[0]:
raw_text = output[0][-1][1]
except: pass
if raw_text:
json_string = None
json_pattern = r"`{3}json\s*(\{.*?\})\s*`{3}"
match = re.search(json_pattern, raw_text, re.DOTALL)
if match: json_string = match.group(1)
else:
s_idx = raw_text.find('{')
e_idx = raw_text.rfind('}')
if s_idx != -1 and e_idx != -1: json_string = raw_text[s_idx:e_idx+1]
if json_string:
data = json.loads(json_string)
break
except Exception as e:
time.sleep(2)
if not data or "script" not in data:
raise ValueError("هوش مصنوعی در نوشتن سناریو با خطا مواجه شد.")
for t in data["script"]:
if "dialogue" in t: t["dialogue"] = re.sub(r'\[.*?\]|\(.*?\)', '', t["dialogue"]).strip()
script_turns = [t for t in data.get("script", []) if str(t.get("dialogue", "")).strip()]
total_turns = len(script_turns)
logging.info(f"✅ سناریو استخراج شد! شامل {total_turns} نوبت گفتگو.")
with tasks_lock:
tasks[task_id] = {'status': 'generating_audio', 'progress': f'در حال تولید صداها (0 از {total_turns} تکمیل شده)'}
audio_results = [None] * total_turns
completed_count = 0
def process_single_chunk(index, turn_data):
raw_speaker_id = str(turn_data.get("speaker_id", "")).strip()
final_speaker_id = raw_speaker_id
if final_speaker_id not in valid_speaker_ids:
found = False
for v_id in valid_speaker_ids:
if final_speaker_id.lower() == v_id.lower():
final_speaker_id = v_id
found = True
break
if not found:
logging.warning(f"⚠️ آیدی گوینده نامعتبر '{raw_speaker_id}' تشخیص داده شد. جایگزین شد با '{default_speaker_id}'.")
final_speaker_id = default_speaker_id
dialogue = turn_data.get("dialogue")
payload = {"text": dialogue, "speaker": final_speaker_id, "temperature": 0.9, "is_custom": False}
idx, audio_seg = call_worker(index, payload)
if audio_seg is None:
raise ValueError(f"خطا در تولید صدای نوبت {index+1}")
return idx, audio_seg
batch_size = 7
batches = [script_turns[i:i + batch_size] for i in range(0, total_turns, batch_size)]
for batch_index, batch in enumerate(batches):
logging.info(f"🔄 شروع پردازش دسته‌ی {batch_index+1} از {len(batches)} (شامل {len(batch)} دیالوگ)...")
with concurrent.futures.ThreadPoolExecutor(max_workers=len(batch)) as executor:
start_idx = batch_index * batch_size
future_to_index = {executor.submit(process_single_chunk, start_idx + i, turn): start_idx + i for i, turn in enumerate(batch)}
for future in concurrent.futures.as_completed(future_to_index):
try:
idx, audio_seg = future.result()
audio_results[idx] = audio_seg
completed_count += 1
with tasks_lock:
tasks[task_id]['progress'] = f'در حال تولید صداها ({completed_count} از {total_turns} تکمیل شده)'
except Exception as exc:
logging.error(f"❌ خطای بحرانی: {str(exc)}")
raise Exception(str(exc))
logging.info(f"✔️ دسته‌ی {batch_index+1} با موفقیت تمام شد.")
logging.info("✂️ تمام قطعات ساخته شد. در حال ادغام (میکس) فایل‌ها...")
combined_audio = AudioSegment.empty()
for seg in audio_results:
if seg:
combined_audio += seg
with tasks_lock:
tasks[task_id] = {'status': 'mixing', 'progress': 'در حال میکس نهایی و ذخیره پادکست...'}
filename = f"full_podcast_{task_id}.mp3"
filepath = os.path.join(CACHE_DIRECTORY, filename)
output_buffer = io.BytesIO()
combined_audio.export(output_buffer, format="mp3", bitrate="192k")
output_buffer.seek(0)
with open(filepath, 'wb') as f:
f.write(output_buffer.read())
logging.info(f"🎉 پادکست نهایی آماده شد! نام فایل: {filename}")
with tasks_lock:
tasks[task_id] = {'status': 'completed', 'filename': filename, 'progress': 'پادکست شما آماده است!'}
except Exception as e:
logging.error(f"❌ پردازش کل پادکست با شکست مواجه شد: {str(e)}")
with tasks_lock:
tasks[task_id] = {'status': 'failed', 'error': str(e)}
# --- VC LOGIC ---
def process_voice_conversion(tts_audio_io, ref_audio_base64):
try:
tts_audio_io.seek(0)
if "," in ref_audio_base64: ref_audio_base64 = ref_audio_base64.split(",")[1]
ref_bytes = base64.b64decode(ref_audio_base64)
files = {
'source_audio': ('source.wav', tts_audio_io, 'audio/wav'),
'ref_audio': ('ref.wav', io.BytesIO(ref_bytes), 'audio/wav')
}
res = requests.post(f"{VC_SPACE_URL}/upload", files=files, headers=STANDARD_HEADERS, timeout=120)
if res.status_code != 200: raise Exception(f"VC Upload Failed: {res.text}")
job_data = res.json()
for _ in range(120):
time.sleep(4)
chk = requests.post(f"{VC_SPACE_URL}/check_status", json=job_data, headers=STANDARD_HEADERS, timeout=30)
if chk.status_code == 200:
stat = chk.json()
if stat.get("status") == "completed":
filename = stat.get("filename")
dl = requests.get(f"{VC_SPACE_URL}/download/{filename}", headers=STANDARD_HEADERS)
if dl.status_code == 200: return io.BytesIO(dl.content)
else: raise Exception("VC Download Failed")
elif stat.get("status") == "failed":
raise Exception(f"VC Remote Failed: {stat.get('detail', 'Unknown error')}")
raise Exception("VC Timeout (Processing took too long)")
except Exception as e:
logging.error(f"VC Error: {e}")
return None
# --- API METHODS FOR FRIENDS (FAST INTEGRATION) ---
def build_fast_scenario(prompt):
spk_text = "- Charon: شهاب (مرد)\n- Zephyr: آوا (زن)"
system_prompt = f"""Act as a Professional Podcast Producer.\nTopic: "{prompt}"\nSpeakers Available:\n{spk_text}\nCRITICAL INSTRUCTION: You must create a VERY LONG, in-depth, and highly detailed podcast script.\nOutput ONLY valid JSON.\nFormat: {{"selected_speakers": ["Charon", "Zephyr"], "script": [{{"speaker_id": "Charon", "dialogue": "..."}}]}}"""
AYA_SPACE_URL = "https://coherelabs-aya-expanse.hf.space/gradio_api"
for attempt in range(5):
try:
session_hash = str(uuid.uuid4())[0:11]
payload = {"data": [system_prompt, [], None, None], "event_data": None, "fn_index": 2, "session_hash": session_hash, "trigger_id": 37}
join_response = requests.post(f"{AYA_SPACE_URL}/queue/join", json=payload, headers=STANDARD_HEADERS, timeout=60)
join_response.raise_for_status()
stream_url = f"{AYA_SPACE_URL}/queue/data?session_hash={session_hash}"
raw_text = None
with requests.get(stream_url, stream=True, headers=STANDARD_HEADERS, timeout=120) as resp:
for line in resp.iter_lines():
if line:
decoded_line = line.decode('utf-8')
if decoded_line.startswith('data: '):
try:
msg_data = json.loads(decoded_line[6:])
if msg_data.get('msg') == 'process_completed':
output = msg_data.get('output', {}).get('data', [])
if output and len(output) > 0 and output[0]:
raw_text = output[0][-1][1]
except: pass
if raw_text:
json_string = None
match = re.search(r"`{3}json\s*(\{.*?\})\s*`{3}", raw_text, re.DOTALL)
if match: json_string = match.group(1)
else:
s_idx = raw_text.find('{')
e_idx = raw_text.rfind('}')
if s_idx != -1 and e_idx != -1: json_string = raw_text[s_idx:e_idx+1]
if json_string:
data = json.loads(json_string)
return data
except Exception as e:
time.sleep(2)
raise ValueError("تولید سناریو با مشکل مواجه شد")
# --- ROUTES ---
@app.route('/')
def index():
return render_template('index.html')
@app.route('/api/fast/tts', methods=['GET', 'POST'])
def fast_tts():
try:
data = request.args if request.method == 'GET' else request.get_json(silent=True) or request.form
text = data.get("text", "").strip()
speaker = data.get("speaker", "Charon")
if not text:
return jsonify({"error": "پارامتر text الزامی است"}), 400
payload = {"text": text, "speaker": speaker, "temperature": 0.9, "is_custom": False}
idx, audio_seg = call_worker(0, payload)
if audio_seg is None:
return jsonify({"error": "خطا در تولید صدا"}), 500
mp3_buffer = io.BytesIO()
audio_seg.export(mp3_buffer, format="mp3", bitrate="192k")
mp3_buffer.seek(0)
return send_file(mp3_buffer, mimetype="audio/mpeg", as_attachment=True, download_name=f"tts_{uuid.uuid4()}.mp3")
except Exception as e:
return jsonify({"error": str(e)}), 500
@app.route('/api/fast/scenario', methods=['GET', 'POST'])
def fast_scenario():
try:
data = request.args if request.method == 'GET' else request.get_json(silent=True) or request.form
prompt = data.get("prompt", "").strip()
if not prompt:
return jsonify({"error": "پارامتر prompt الزامی است"}), 400
data = build_fast_scenario(prompt)
return jsonify(data)
except Exception as e:
return jsonify({"error": str(e)}), 500
@app.route('/api/fast/podcast', methods=['GET', 'POST'])
def fast_podcast():
try:
data = request.args if request.method == 'GET' else request.get_json(silent=True) or request.form
prompt = data.get("prompt", "").strip()
if not prompt:
return jsonify({"error": "پارامتر prompt الزامی است"}), 400
script_data = build_fast_scenario(prompt)
script_turns = script_data.get("script", [])
valid_speaker_ids = ["Charon", "Zephyr"]
audio_results = [None] * len(script_turns)
def process_chunk(index, turn_data):
spk = turn_data.get("speaker_id", "Charon")
if spk not in valid_speaker_ids: spk = "Charon"
payload = {"text": turn_data.get("dialogue", ""), "speaker": spk, "temperature": 0.9, "is_custom": False}
idx, audio_seg = call_worker(index, payload)
if audio_seg is None: raise ValueError("خطا")
return idx, audio_seg
with concurrent.futures.ThreadPoolExecutor(max_workers=5) as executor:
future_to_idx = {executor.submit(process_chunk, i, t): i for i, t in enumerate(script_turns)}
for future in concurrent.futures.as_completed(future_to_idx):
idx, audio_seg = future.result()
audio_results[idx] = audio_seg
combined_audio = AudioSegment.empty()
for seg in audio_results:
if seg: combined_audio += seg
mp3_buffer = io.BytesIO()
combined_audio.export(mp3_buffer, format="mp3", bitrate="192k")
mp3_buffer.seek(0)
return send_file(mp3_buffer, mimetype="audio/mpeg", as_attachment=True, download_name=f"podcast_{uuid.uuid4()}.mp3")
except Exception as e:
return jsonify({"error": str(e)}), 500
@app.route('/api/check-credit', methods=['POST'])
def check_credit():
data = request.get_json()
fingerprint = data.get('fingerprint')
if not fingerprint: return jsonify({"status": "error"}), 400
ip = get_user_ip()
now = time.time()
week_ago = now - (7*24*60*60)
try:
with get_db_connection() as conn:
user = get_user_record(conn, fingerprint, ip)
if user:
if user['week_start'] < week_ago:
conn.execute('UPDATE users SET count = 0, week_start = ? WHERE fingerprint = ?', (now, user['fingerprint']))
conn.commit()
user['count'] = 0
user['week_start'] = now
remaining = USAGE_LIMIT - user['count']
limit_reached = remaining <= 0
reset_ts = user['week_start'] + (7*24*60*60)
return jsonify({"credits_remaining": max(0, remaining), "limit_reached": limit_reached, "reset_timestamp": reset_ts})
else:
return jsonify({"credits_remaining": USAGE_LIMIT, "limit_reached": False, "reset_timestamp": 0})
except Exception as e:
logging.error(f"Error checking credit: {e}")
return jsonify({"status": "error"}), 500
@app.route('/api/use-credit', methods=['POST'])
def use_credit():
data = request.get_json()
fingerprint = data.get('fingerprint')
if not fingerprint: return jsonify({"status": "error"}), 400
ip = get_user_ip()
now = time.time()
week_ago = now - (7*24*60*60)
try:
with get_db_connection() as conn:
user = get_user_record(conn, fingerprint, ip)
if user:
if user['week_start'] < week_ago:
user['count'] = 0
user['week_start'] = now
if user['count'] >= USAGE_LIMIT:
return jsonify({"status": "limit"}), 429
user['count'] += 1
ips = json.loads(user['ips'])
if ip not in ips:
ips.append(ip)
conn.execute('UPDATE users SET count = ?, week_start = ?, ips = ? WHERE fingerprint = ?',
(user['count'], user['week_start'], json.dumps(ips), user['fingerprint']))
conn.commit()
return jsonify({"status": "success", "credits_remaining": USAGE_LIMIT - user['count']})
else:
conn.execute('INSERT INTO users (fingerprint, ips, count, week_start) VALUES (?, ?, ?, ?)',
(fingerprint, json.dumps([ip]), 1, now))
conn.commit()
return jsonify({"status": "success", "credits_remaining": USAGE_LIMIT - 1})
except Exception as e:
logging.error(f"Error using credit: {e}")
return jsonify({"status": "error"}), 500
@app.route('/api/create-full-podcast', methods=['POST'])
def create_full_podcast():
try:
data = request.get_json()
prompt = data.get('prompt')
speakers = data.get('available_speakers')
if not prompt or not speakers: return jsonify({"error": "Bad request"}), 400
task_id = str(uuid.uuid4())
with tasks_lock: tasks[task_id] = {'status': 'pending'}
safety = [{"category": c, "threshold": "BLOCK_NONE"} for c in ["HARM_CATEGORY_HARASSMENT", "HARM_CATEGORY_HATE_SPEECH", "HARM_CATEGORY_SEXUALLY_EXPLICIT", "HARM_CATEGORY_DANGEROUS_CONTENT"]]
spk_text = "\n".join([f"- {s['id']}: {s['name']}" for s in speakers])
sys_prompt = f"""Act as a Professional Podcast Producer.\nTopic: "{prompt}"\nSpeakers Available:\n{spk_text}\nCRITICAL INSTRUCTIONS:\n1. Create a VERY LONG, in-depth podcast script.\n2. Keep EVERY dialogue line SHORT. Break long speeches into multiple consecutive turns for the same speaker.\n3. NO stage directions, NO emojis, NO brackets. Plain text ONLY.\n\nOutput ONLY valid JSON.\nFormat: {{"selected_speakers": ["id1", "id2"], "script": [{{"speaker_id": "id1", "dialogue": "..."}}]}}"""
threading.Thread(target=generate_podcast_in_background, args=(task_id, sys_prompt, safety)).start()
return jsonify({"task_id": task_id}), 202
except Exception as e:
return jsonify({"error": str(e)}), 500
@app.route('/api/podcast-status/<task_id>', methods=['GET'])
def podcast_status(task_id):
with tasks_lock:
return jsonify(tasks.get(task_id, {'status': 'not_found'})), 200
@app.route('/api/auto-podcast', methods=['POST'])
def auto_podcast():
try:
data = request.get_json()
prompt = data.get('prompt')
speakers = data.get('available_speakers')
if not prompt or not speakers: return jsonify({"error": "Bad request"}), 400
task_id = str(uuid.uuid4())
with tasks_lock: tasks[task_id] = {'status': 'pending', 'progress': 'در حال ورود به صف پردازش...'}
threading.Thread(target=generate_full_podcast_audio_background, args=(task_id, prompt, speakers)).start()
return jsonify({"task_id": task_id}), 202
except Exception as e:
return jsonify({"error": str(e)}), 500
@app.route('/api/auto-podcast-status/<task_id>', methods=['GET'])
def auto_podcast_status(task_id):
with tasks_lock:
return jsonify(tasks.get(task_id, {'status': 'not_found'})), 200
@app.route('/api/download-podcast/<filename>', methods=['GET'])
def download_podcast(filename):
filepath = os.path.join(CACHE_DIRECTORY, filename)
if os.path.exists(filepath):
return send_file(filepath, mimetype="audio/mpeg", as_attachment=True)
return jsonify({"error": "File not found"}), 404
@app.route('/api/generate', methods=['POST'])
def generate_audio_route():
try:
data = request.get_json()
if not data: return jsonify({"error": "No data"}), 400
text = data.get("text", "")
speaker = data.get("speaker")
temperature = data.get("temperature", 0.9)
ref_base64 = data.get("ref_audio_base64")
if not text: return jsonify({"error": "Text empty"}), 400
is_custom = bool(speaker.startswith("custom_") and ref_base64)
payload = {"text": text, "speaker": speaker, "temperature": temperature, "is_custom": is_custom}
idx, audio_seg = call_worker(0, payload)
if audio_seg is None: return jsonify({"error": "Worker generation failed"}), 503
if is_custom:
logging.info("Starting Custom VC...")
# برای اسپیس تغییر صدا (VC) اول فایل موقت wav میسازیم
wav_buffer = io.BytesIO()
audio_seg.export(wav_buffer, format="wav")
wav_buffer.seek(0)
vc_out = process_voice_conversion(wav_buffer, ref_base64)
if vc_out:
# خروجی نهایی VC رو به صورت قطعی به MP3 تبدیل میکنیم
vc_audio = AudioSegment.from_file(vc_out)
mp3_buffer = io.BytesIO()
vc_audio.export(mp3_buffer, format="mp3", bitrate="192k")
mp3_buffer.seek(0)
return send_file(mp3_buffer, mimetype="audio/mpeg", as_attachment=True, download_name=f"vc_{uuid.uuid4()}.mp3")
else:
return jsonify({"error": "Voice Conversion failed"}), 500
# تبدیل مستقیم صدای تولید شده معمولی به MP3
mp3_buffer = io.BytesIO()
audio_seg.export(mp3_buffer, format="mp3", bitrate="192k")
mp3_buffer.seek(0)
return send_file(mp3_buffer, mimetype="audio/mpeg", as_attachment=True, download_name=f"gen_{uuid.uuid4()}.mp3")
except Exception as e:
logging.error(f"Generate route error: {e}")
return jsonify({"error": str(e)}), 500
# --- STARTUP ---
init_db()
threading.Thread(target=background_cleanup, daemon=True).start()
if __name__ == '__main__':
port = int(os.environ.get('PORT', 7860))
app.run(host='0.0.0.0', port=port)