import os
import base64
import uuid
import gradio as gr
from openai import OpenAI
from speechify import Speechify
from dotenv import load_dotenv

# Detect Hugging Face environment
RUNNING_IN_SPACES = os.getenv("SYSTEM") == "spaces"

# Load API keys
if not RUNNING_IN_SPACES:
    load_dotenv()
openai_api_key = os.getenv("OPENAI_API_KEY")
speechify_api_key = os.getenv("SPEECHIFY_API_KEY")

# Sanity check (but don't print full keys)
print(f"✅ OPENAI_API_KEY loaded: {'✅' if openai_api_key else '❌ MISSING'}")
print(f"✅ SPEECHIFY_API_KEY loaded: {'✅' if speechify_api_key else '❌ MISSING'}")

# Initialize clients
openai_client = OpenAI(api_key=openai_api_key)
speechify_client = Speechify(token=speechify_api_key)

# Voice config
language_config = {
    "Portuguese": {
        "voice_id": "joao",
        "language": "pt-PT",
        "model": "simba-multilingual",
        "audio_format": "mp3"
    },
    "French": {
        "voice_id": "leo",
        "language": "fr-FR",
        "model": "simba-multilingual",
        "audio_format": "mp3"
    },
    "Spanish": {
        "voice_id": "danna-sofia",
        "language": "es-MX",
        "model": "simba-multilingual",
        "audio_format": "mp3"
    },
}

def chat_and_speak(user_input, language_choice):
    gpt_response = ""
    audio_output_path = None
    try:
        if not user_input or not user_input.strip():
            return None, "Please enter some text to process."

        print(f"🧠 User input: {user_input}")
        print(f"🗣️ Language choice: {language_choice}")

        # Step 1: Get GPT response
        system_message = f"You are a friendly {language_choice} language tutor. Respond only in {language_choice}."
        completion = openai_client.chat.completions.create(
            model="gpt-4",
            messages=[
                {"role": "system", "content": system_message},
                {"role": "user", "content": user_input}
            ]
        )
        gpt_response = completion.choices[0].message.content
        print(f"💬 GPT response: {gpt_response}")

        # Step 2: Use Speechify to generate audio
        config = language_config.get(language_choice)
        if not config:
            error_msg = f"⚠️ Language '{language_choice}' not supported."
            print(error_msg)
            return None, f"{gpt_response}\n\n{error_msg}"

        tts_response = speechify_client.tts.audio.speech(
            input=gpt_response,
            voice_id=config["voice_id"],
            model=config["model"],
            audio_format=config["audio_format"]
        )

        if hasattr(tts_response, "audio_data") and isinstance(tts_response.audio_data, str) and tts_response.audio_data:
            try:
                audio_bytes = base64.b64decode(tts_response.audio_data)
                output_dir = "/tmp" if RUNNING_IN_SPACES else "speech_files"
                os.makedirs(output_dir, exist_ok=True)
                audio_output_path = os.path.join(output_dir, f"speech_{uuid.uuid4().hex}.mp3")
                with open(audio_output_path, "wb") as f:
                    f.write(audio_bytes)
            except Exception as audio_err:
                print(f"🔥 Error processing audio data: {audio_err}")
                return None, f"{gpt_response}\n\n⚠️ Error saving audio: {audio_err}"
        else:
            print("⚠️ No audio data received from Speechify or audio_data is not a string.")
            return None, f"{gpt_response}\n\n⚠️ No audio data received from Speechify."

        return audio_output_path, gpt_response

    except Exception as e:
        print(f"🔥 An unexpected error occurred: {e}")
        error_message = f"⚠️ An unexpected error occurred: {e}"
        if gpt_response:
            return None, f"{gpt_response}\n\n{error_message}"
        return None, error_message

with open("custom.css") as f:
    custom_css = f.read()


with gr.Blocks(css=custom_css) as demo:
    gr.HTML(
        '<div class="custom-bar"><span class="custom-bar-title">Language Tutor</span></div>'
    )  
    with gr.Column(elem_classes="main-card"):
        with gr.Row():
            with gr.Column():
                user_input = gr.Textbox(label="Type in whatever language you prefer", placeholder="Type here...", lines=4)
                language_choice = gr.Dropdown(
                    choices=["Portuguese", "French", "Spanish"],
                    value="Portuguese",
                    label="Language"
                )
                submit_btn = gr.Button("Submit")
            with gr.Column():
                audio_output = gr.Audio(label="Audio Playback", type="filepath", autoplay=True)
                gpt_output = gr.Textbox(label="The Response")

        submit_btn.click(
            fn=chat_and_speak,
            inputs=[user_input, language_choice],
            outputs=[audio_output, gpt_output]
        )

if __name__ == "__main__":
    demo.launch()