voice_pipeline.py

import os
import time
import json
import base64
import requests
import tempfile
import threading
import queue
import winsound

TRANSCRIPT_PATH = r"C:\Users\nasko\.gemini\antigravity\brain\695897cf-1c58-4886-a686-e9d8c406ebef\.system_generated\logs\transcript.jsonl"
API_URL = "http://localhost:8000/api/v1/synthesize/stream"

audio_queue = queue.Queue()

def player_worker():
    """
    Взима готови WAV файлове от опашката и ги пуска.
    """
    while True:
        file_path = audio_queue.get()
        if file_path is None: break
        
        print(f"🔊 Възпроизвеждам от API...")
        winsound.PlaySound(file_path, winsound.SND_FILENAME)
        
        try:
            os.remove(file_path)
        except OSError:
            pass
            
        audio_queue.task_done()

def process_text(text: str):
    """
    Изпраща текста към API-то и чака за стрийминг на аудио парчета.
    """
    print(f"📡 Изпращане към API: {text[:50]}...")
    try:
        response = requests.post(API_URL, json={
            "text": text,
            "voice_style": "F5",
            "speed": 1.6
        }, stream=True)
        
        if response.status_code != 200:
            print(f"Грешка от API: {response.status_code} - {response.text}")
            return
            
        for line in response.iter_lines():
            if line:
                data = json.loads(line)
                if "error" in data:
                    print(f"API Грешка: {data['error']}")
                    continue
                    
                chunk_index = data.get("chunk_index")
                audio_base64 = data.get("audio_base64")
                
                if audio_base64:
                    audio_bytes = base64.b64decode(audio_base64)
                    
                    # Записваме временно файла и го пускаме в опашката
                    fd, file_path = tempfile.mkstemp(suffix=f"_chunk_{chunk_index}.wav")
                    os.close(fd)
                    
                    with open(file_path, "wb") as f:
                        f.write(audio_bytes)
                        
                    audio_queue.put(file_path)
    except requests.exceptions.ConnectionError:
        print("Не мога да се свържа с API-то! Увери се, че `api.py` работи на порт 8000.")
    except Exception as e:
        print(f"Грешка при комуникация с API: {e}")

def tail_file():
    """
    Следи чата (transcript.jsonl) за нови съобщения от модела.
    """
    if not os.path.exists(TRANSCRIPT_PATH):
        print(f"Файлът не съществува: {TRANSCRIPT_PATH}")
        return
        
    with open(TRANSCRIPT_PATH, "r", encoding="utf-8") as f:
        f.seek(0, 2) 
        
        while True:
            line = f.readline()
            if not line:
                time.sleep(0.5)
                continue
            
            try:
                data = json.loads(line)
                if data.get("source") == "MODEL" and data.get("type") in ["PLANNER_RESPONSE", "GENERIC"]:
                    full_text = data.get("content", "")
                    if full_text and not full_text.startswith("Created At:"):
                        print("\n📝 Получен нов текст от чата.")
                        process_text(full_text)
            except Exception as e:
                pass

if __name__ == "__main__":
    t_play = threading.Thread(target=player_worker, daemon=True)
    t_play.start()
    
    print("Ani Voice Client слуша за съобщения и чака API-то...")
    tail_file()