app.py

# ===================================================================================
# 1. SETUP AND IMPORTS
# ===================================================================================
from __future__ import annotations
import os
import requests
import base64
import struct
import re
import textwrap
import uuid
from typing import List, Dict, Tuple, Generator

# Make sure Gradio analytics is off (so we don't need pandas 2.x)
os.environ.setdefault("GRADIO_ANALYTICS_ENABLED", "False")

# --- Load .env early (for HF_TOKEN / SECRET_TOKEN) ---
from dotenv import load_dotenv
load_dotenv()

# --- Hugging Face Spaces & ZeroGPU ---
try:
    import spaces  # Required for ZeroGPU on HF
except Exception:
    # Allow local runs without the spaces package
    class _SpacesShim:
        def GPU(self, *args, **kwargs):
            def _wrap(fn):
                return fn
            return _wrap
    spaces = _SpacesShim()

import gradio as gr

# --- Core ML & Data Libraries ---
import torch
import numpy as np
from huggingface_hub import HfApi, hf_hub_download
from llama_cpp import Llama

# --- TTS Libraries ---
from TTS.tts.configs.xtts_config import XttsConfig
from TTS.tts.models.xtts import Xtts
from TTS.utils.manage import ModelManager
from TTS.utils.generic_utils import get_user_data_dir

# --- Text & Audio Processing ---
import nltk
import langid
import emoji
import noisereduce as nr

# ===================================================================================
# 2. GLOBAL CONFIGURATION & HELPER FUNCTIONS
# ===================================================================================

# Download NLTK data (punkt)
nltk.download("punkt", quiet=True)

os.environ["COQUI_TOS_AGREED"] = "1"

# Cached models
tts_model: Xtts | None = None
llm_model: Llama | None = None

# Configuration
HF_TOKEN = os.environ.get("HF_TOKEN")
api = HfApi(token=HF_TOKEN) if HF_TOKEN else None
repo_id = "ruslanmv/ai-story-server"
SECRET_TOKEN = os.getenv("SECRET_TOKEN", "secret")
SENTENCE_SPLIT_LENGTH = 250
LLM_STOP_WORDS = ["</s>", "<|user|>", "/s>"]

# System prompts and roles
default_system_message = (
    "You're a storyteller crafting a short tale for young listeners. Keep sentences short and simple. "
    "Use narrative style only, without lists or complex words. Type numbers as words (e.g., 'ten')."
)
system_message = os.environ.get("SYSTEM_MESSAGE", default_system_message)
ROLES = ["Cloée", "Julian", "Pirate", "Thera"]
ROLE_PROMPTS = {role: system_message for role in ROLES}
ROLE_PROMPTS["Pirate"] = (
    "You are AI Beard, a pirate. Craft your response from his first-person perspective. "
    "Keep answers short, as if in a real conversation. Only provide the words AI Beard would speak."
)

# --- Audio helpers ---
def pcm_to_wav(pcm_data: bytes, sample_rate: int = 24000, channels: int = 1, bit_depth: int = 16) -> bytes:
    if pcm_data.startswith(b"RIFF"):
        return pcm_data
    chunk_size = 36 + len(pcm_data)
    header = struct.pack(
        "<4sI4s4sIHHIIHH4sI",
        b"RIFF", chunk_size, b"WAVE", b"fmt ",
        16, 1, channels, sample_rate,
        sample_rate * channels * bit_depth // 8,
        channels * bit_depth // 8, bit_depth,
        b"data", len(pcm_data)
    )
    return header + pcm_data

def split_sentences(text: str, max_len: int) -> List[str]:
    sentences = nltk.sent_tokenize(text)
    chunks: List[str] = []
    for sent in sentences:
        if len(sent) > max_len:
            chunks.extend(textwrap.wrap(sent, max_len, break_long_words=True))
        else:
            chunks.append(sent)
    return chunks

def format_prompt_zephyr(message: str, history: List[Tuple[str, str | None]], system_message: str) -> str:
    prompt = f"<|system|>\n{system_message}</s>"
    for user_prompt, bot_response in history:
        if bot_response:
            prompt += f"<|user|>\n{user_prompt}</s><|assistant|>\n{bot_response}</s>"
    prompt += f"<|user|>\n{message}</s><|assistant|>"
    return prompt

# ===================================================================================
# 3. CORE AI FUNCTIONS (Model Loading & Inference)
# ===================================================================================

def _load_xtts(device: str) -> Xtts:
    print("Loading Coqui XTTS V2 model (first run)...")
    model_name = "tts_models/multilingual/multi-dataset/xtts_v2"
    ModelManager().download_model(model_name)
    model_path = os.path.join(get_user_data_dir("tts"), model_name.replace("/", "--"))

    config = XttsConfig()
    config.load_json(os.path.join(model_path, "config.json"))
    model = Xtts.init_from_config(config)
    # NOTE: deepspeed not installed; keep False for Spaces
    model.load_checkpoint(
        config,
        checkpoint_path=os.path.join(model_path, "model.pth"),
        vocab_path=os.path.join(model_path, "vocab.json"),
        eval=True,
        use_deepspeed=False,
    )
    model.to(device)
    print("XTTS model loaded.")
    return model

def _load_llama() -> Llama:
    print("Loading LLM (Zephyr) (first run)...")
    zephyr_model_path = hf_hub_download(
        repo_id="TheBloke/zephyr-7B-beta-GGUF",
        filename="zephyr-7b-beta.Q5_K_M.gguf"
    )
    # Try GPU offload if available, else CPU
    for n_gpu_layers in (-1, 0):
        try:
            llm = Llama(
                model_path=zephyr_model_path,
                n_gpu_layers=n_gpu_layers,
                n_ctx=4096,
                n_batch=512,
                verbose=False
            )
            if n_gpu_layers == -1:
                print("LLM loaded with GPU offload.")
            else:
                print("LLM loaded (CPU).")
            return llm
        except Exception as e:
            print(f"LLM init with n_gpu_layers={n_gpu_layers} failed: {e}")
    raise RuntimeError("Failed to initialize Llama model.")

def load_models() -> Tuple[Xtts, Llama]:
    global tts_model, llm_model
    device = "cuda" if torch.cuda.is_available() else "cpu"
    if tts_model is None:
        tts_model = _load_xtts(device)
    if llm_model is None:
        llm_model = _load_llama()
    return tts_model, llm_model

def generate_text_stream(llm_instance: Llama, prompt: str,
                         history: List[Tuple[str, str | None]],
                         system_message: str) -> Generator[str, None, None]:
    formatted_prompt = format_prompt_zephyr(prompt, history, system_message)
    stream = llm_instance(
        formatted_prompt,
        temperature=0.7,
        max_tokens=512,
        top_p=0.95,
        stop=LLM_STOP_WORDS,
        stream=True
    )
    for response in stream:
        ch = response["choices"][0]["text"]
        # Guard against control tokens & isolated emoji artefacts
        try:
            is_single_emoji = (len(ch) == 1 and emoji.is_emoji(ch))  # emoji>=2.x
        except Exception:
            is_single_emoji = False
        if "<|user|>" in ch or is_single_emoji:
            continue
        yield ch

def generate_audio_stream(tts_instance: Xtts, text: str, language: str,
                          latents: Tuple[np.ndarray, np.ndarray]) -> Generator[bytes, None, None]:
    gpt_cond_latent, speaker_embedding = latents
    try:
        for chunk in tts_instance.inference_stream(
            text=text,
            language=language,
            gpt_cond_latent=gpt_cond_latent,
            speaker_embedding=speaker_embedding,
            temperature=0.85,
        ):
            if chunk is not None:
                yield chunk.detach().cpu().numpy().squeeze().tobytes()
    except RuntimeError as e:
        print(f"Error during TTS inference: {e}")
        # Soft-restart if GPU went bad and we can talk to the HF API
        if "device-side assert" in str(e) and api:
            gr.Warning("Critical GPU error. Attempting to restart the Space...")
            try:
                api.restart_space(repo_id=repo_id)
            except Exception:
                pass

# ===================================================================================
# 4. MAIN GRADIO FUNCTION (Decorated for ZeroGPU)
# ===================================================================================

@spaces.GPU(duration=120)  # Request GPU for 120 seconds
def generate_story_and_speech(secret_token_input: str, input_text: str, chatbot_role: str) -> List[Dict[str, str]]:
    if secret_token_input != SECRET_TOKEN:
        raise gr.Error("Invalid secret token provided.")
    if not input_text:
        return []

    # Load models
    tts, llm = load_models()

    # Pre-compute voice latents
    latent_map: Dict[str, Tuple[np.ndarray, np.ndarray]] = {}
    for role, filename in [
        ("Cloée", "cloee-1.wav"),
        ("Julian", "julian-bedtime-style-1.wav"),
        ("Pirate", "pirate_by_coqui.wav"),
        ("Thera", "thera-1.wav"),
    ]:
        path = os.path.join("voices", filename)
        latent_map[role] = tts.get_conditioning_latents(
            audio_path=path, gpt_cond_len=30, max_ref_length=60
        )

    # Generate story text
    history: List[Tuple[str, str | None]] = [(input_text, None)]
    full_story_text = "".join(
        generate_text_stream(llm, history[-1][0], history[:-1], system_message=ROLE_PROMPTS[chatbot_role])
    ).strip()

    if not full_story_text:
        return []

    # Tokenize into shorter sentences for TTS
    sentences = split_sentences(full_story_text, SENTENCE_SPLIT_LENGTH)
    lang = langid.classify(sentences[0])[0] if sentences else "en"

    results: List[Dict[str, str]] = []
    for sentence in sentences:
        if not any(c.isalnum() for c in sentence):
            continue

        audio_chunks = generate_audio_stream(tts, sentence, lang, latent_map[chatbot_role])
        pcm_data = b"".join(chunk for chunk in audio_chunks if chunk)

        # Optional noise reduction (best-effort)
        try:
            data_s16 = np.frombuffer(pcm_data, dtype=np.int16)
            if data_s16.size > 0:
                float_data = data_s16.astype(np.float32) / 32767.0
                reduced = nr.reduce_noise(y=float_data, sr=24000)
                final_pcm = (reduced * 32767).astype(np.int16).tobytes()
            else:
                final_pcm = pcm_data
        except Exception:
            final_pcm = pcm_data

        b64_wav = base64.b64encode(pcm_to_wav(final_pcm)).decode("utf-8")
        results.append({"text": sentence, "audio": b64_wav})

    return results

# ===================================================================================
# 5. GRADIO INTERFACE LAUNCH
# ===================================================================================

# Download voice files on startup
print("Downloading voice files...")
file_names = ["cloee-1.wav", "julian-bedtime-style-1.wav", "pirate_by_coqui.wav", "thera-1.wav"]
base_url = "https://raw.githubusercontent.com/ruslanmv/ai-story-server/main/voices/"
os.makedirs("voices", exist_ok=True)
for name in file_names:
    dst = os.path.join("voices", name)
    if not os.path.exists(dst):
        try:
            resp = requests.get(base_url + name, timeout=30)
            resp.raise_for_status()
            with open(dst, "wb") as f:
                f.write(resp.content)
        except Exception as e:
            print(f"Failed to download {name}: {e}")

# Define the Gradio Interface
demo = gr.Interface(
    fn=generate_story_and_speech,
    inputs=[
        gr.Textbox(label="Secret Token", type="password", value=SECRET_TOKEN),
        gr.Textbox(placeholder="What should the story be about?", label="Story Prompt"),
        gr.Dropdown(choices=ROLES, label="Select a Storyteller", value="Cloée"),
    ],
    outputs=gr.JSON(label="Story and Audio Output"),
    title="AI Storyteller with ZeroGPU",
    description="Enter a prompt to generate a short story with voice narration using on-demand GPU.",
    allow_flagging="never",
    analytics_enabled=False,  # <- keep analytics off to avoid pandas 2.x requirement
)

if __name__ == "__main__":
    # For Spaces or Docker, these defaults are handy; adjust as needed.
    demo.queue().launch(server_name="0.0.0.0", server_port=int(os.getenv("PORT", "7860")))