voice_agent_gui.py

import os
import gradio as gr
import tempfile
import torchaudio
from dotenv import load_dotenv
from tortoise.api import TextToSpeech
from tortoise.utils.audio import load_voice
from openai import OpenAI

# === Debug and environment setup ===
print(f"Current working directory: {os.getcwd()}")
load_dotenv()  # By default, looks for .env in current working directory

# === Initialize OpenAI client with better error handling ===
api_key = os.getenv("OPENAI_API_KEY")
if not api_key:
    raise ValueError("❌ OPENAI_API_KEY not found in your .env file.")
print(f"API key found: {'Yes' if api_key else 'No'}")
print(f"API key loaded: {api_key[:4]}...{api_key[-4:] if len(api_key) > 8 else ''}")

# Initialize with timeout to prevent hanging requests
client = OpenAI(
    api_key=api_key,
    timeout=60.0  # 60 second timeout
)

# === Initialize Tortoise TTS ===
print("Initializing Tortoise TTS...")
tts = TextToSpeech()
voice_samples, conditioning_latents = load_voice("train_dotrice")
print("TTS initialized successfully!")

# === Ask GPT-4o with improved error handling ===
def ask_gpt(prompt: str) -> str:
    try:
        print(f"Sending request to GPT-4o: {prompt[:30]}...")
        response = client.chat.completions.create(
            model="gpt-4o",
            messages=[{"role": "user", "content": prompt}],
            temperature=0.7,
            max_tokens=300
        )
        return response.choices[0].message.content
    except Exception as e:
        error_type = type(e).__name__
        error_message = str(e)
        print(f"[GPT-4 ERROR] {error_type}: {error_message}")
        
        # Provide more helpful error messages
        if "api_key" in error_message.lower():
            return "[GPT-4 ERROR] API key issue: Check that your API key is valid and properly formatted in the .env file."
        elif "rate limit" in error_message.lower():
            return "[GPT-4 ERROR] Rate limit exceeded: Please wait a moment before trying again."
        elif "connect" in error_message.lower():
            return "[GPT-4 ERROR] Connection error: Check your internet connection and ensure OpenAI's API is accessible."
        else:
            return f"[GPT-4 ERROR] {error_type}: {error_message}"

# === Generate TTS Audio ===
def synthesize(text: str) -> str:
    try:
        print(f"Synthesizing speech for: {text[:30]}...")
        audio = tts.tts_with_preset(
            text=text,
            voice_samples=voice_samples,
            conditioning_latents=conditioning_latents,
            preset="fast"
        )
        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp:
            torchaudio.save(tmp.name, audio.squeeze(0).cpu(), 24000)
            print(f"Audio saved to temporary file: {tmp.name}")
            return tmp.name
    except Exception as e:
        print(f"[TTS ERROR] {type(e).__name__}: {str(e)}")
        return None

# === Unified Agent Logic with Enhanced Error Handling ===
def run_agent(audio_input, text_input):
    try:
        if audio_input:
            return "🧠 Voice transcription not implemented yet.", None

        if text_input and text_input.strip():
            cleaned_input = text_input.strip()
            if len(cleaned_input) < 2:
                return "⚠️ Please enter more meaningful text.", None

            print("Processing text input...")
            gpt_reply = ask_gpt(cleaned_input)
            if gpt_reply.startswith("[GPT-4 ERROR]"):
                return gpt_reply, None
                
            audio_path = synthesize(gpt_reply)
            if audio_path is None:
                return gpt_reply + "\n\n[TTS ERROR] Failed to generate audio.", None
                
            return gpt_reply, audio_path

        return "⚠️ Please enter a message or audio input.", None
    except Exception as e:
        print(f"[AGENT ERROR] {type(e).__name__}: {str(e)}")
        return f"⚠️ An unexpected error occurred: {type(e).__name__}: {str(e)}", None

# === Gradio UI ===
with gr.Blocks() as demo:
    gr.Markdown("## 🧠 GPT-4o Voice Agent + Tortoise TTS")
    gr.Markdown("*Ensure your `.env` file with OPENAI_API_KEY is in the correct directory*")

    with gr.Row():
        mic_input = gr.Audio(
            label="🎙️ Mic Input (WAV format, not yet active)",
            type="filepath",
            format="wav"
        )
        text_input = gr.Textbox(
            lines=2,
            placeholder="Ask anything here...",
            label="💬 Text Input"
        )

    run_btn = gr.Button("🧠 Ask GPT-4o")
    gpt_output = gr.Textbox(label="🧠 GPT-4o Response")
    audio_output = gr.Audio(label="🔊 Spoken Response", autoplay=True)

    run_btn.click(
        fn=run_agent,
        inputs=[mic_input, text_input],
        outputs=[gpt_output, audio_output]
    )

# Launch for local + mobile access
print("Launching Gradio interface...")
demo.launch(share=True, server_name="0.0.0.0", server_port=7860, debug=True)