app.py

import gradio as gr
import google.generativeai as genai
from elevenlabs.client import ElevenLabs
import os
import json
from pypdf import PdfReader
from dotenv import load_dotenv

# 1️⃣ Load API Keys
load_dotenv()
GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
ELEVEN_API_KEY = os.getenv("ELEVEN_API_KEY")

if GEMINI_API_KEY:
    genai.configure(api_key=GEMINI_API_KEY)
if ELEVEN_API_KEY:
    client = ElevenLabs(api_key=ELEVEN_API_KEY)

# 2️⃣ Podcast State
class PodcastState:
    def __init__(self):
        self.script = []
        self.current_index = 0
        self.persona = "Serious Academic"
        self.full_text = ""

state = PodcastState()

# 3️⃣ PDF Extraction
def extract_text_from_pdf(pdf_file):
    try:
        reader = PdfReader(pdf_file)
        text = ""
        for page in reader.pages[:5]:
            page_text = page.extract_text()
            if page_text:
                text += page_text + "\n"
        return text
    except Exception as e:
        return f"Error reading PDF: {e}"

# 4️⃣ Generate Script
def generate_script(pdf_file, persona_style):
    if not pdf_file:
        return "⚠️ Upload PDF first.", {}

    if not GEMINI_API_KEY or not ELEVEN_API_KEY:
        return "⚠️ API Keys missing in HF Secrets!", {}

    pdf_text = extract_text_from_pdf(pdf_file)
    state.full_text = pdf_text
    state.persona = persona_style

    prompts = {
        "Serious Academic": "Two professors discussing the paper. Tone: Intellectual.",
        "Gossip Columnist": "Two gossip hosts reacting dramatically.",
    }

    system_prompt = f"""
    {prompts.get(persona_style)}

    Based on the paper:
    "{pdf_text[:4000]}..."

    Generate a short 4-turn dialogue script.
    Return JSON ONLY:
    [
        {{"speaker":"Host A","text":"..."}},
        {{"speaker":"Host B","text":"..."}}
    ]
    """

    try:
        model = genai.GenerativeModel("gemini-2.0-flash")
        response = model.generate_content(system_prompt)
        clean_json = response.text.replace("```json", "").replace("```", "").strip()
        script = json.loads(clean_json)
        state.script = script
        state.current_index = 0
        return "✅ Script ready!", script
    except Exception as e:
        return f"Error generating script: {e}", {}

# 5️⃣ Play next line
def play_next_chunk():
    if state.current_index >= len(state.script):
        return None, "🎉 Podcast finished."

    line = state.script[state.current_index]
    voice_id = "nPczCjz82tPNOwVbpGE2"
    if state.persona == "Gossip Columnist" and line["speaker"] == "Host B":
        voice_id = "21m00Tcm4TlvDq8ikWAM"

    try:
        audio_stream = client.generate(
            text=line["text"],
            voice=voice_id,
            model="eleven_monolingual_v1"
        )

        save_path = f"temp_{state.current_index}.mp3"
        with open(save_path, "wb") as f:
            for chunk in audio_stream:
                f.write(chunk)

        state.current_index += 1
        return save_path, f"{line['speaker']}: {line['text']}"
    except Exception as e:
        return None, f"Audio error: {e}"

# 6️⃣ Interrupt & Ask
def interrupt_and_ask(question):
    if not state.full_text:
        return None, "Upload a PDF first."

    model = genai.GenerativeModel("gemini-2.0-flash")
    prompt = f"""
    Persona: {state.persona}
    Context: {state.full_text[:1000]}
    User Question: {question}

    Answer briefly, then say "Anyway, back to the paper..."
    """
    try:
        response = model.generate_content(prompt)
        answer = response.text
        audio_stream = client.generate(
            text=answer,
            voice="nPczCjz82tPNOwVbpGE2",
            model="eleven_monolingual_v1"
        )
        save_path = "interrupt.mp3"
        with open(save_path, "wb") as f:
            for chunk in audio_stream:
                f.write(chunk)
        return save_path, answer
    except Exception as e:
        return None, f"Error: {e}"

# 7️⃣ Build Gradio 5.7 UI
with gr.Blocks() as demo:
    gr.Markdown("# 🎧 PodQuery — Research Paper Podcast Generator")

    with gr.Row():
        with gr.Column():
            pdf_input = gr.File(label="Upload PDF", file_types=[".pdf"])
            persona = gr.Dropdown(
                ["Serious Academic", "Gossip Columnist"],
                value="Serious Academic",
                label="Persona Style"
            )
            btn_gen = gr.Button("Generate Podcast Script")
            status = gr.Textbox(label="Status")
            script_display = gr.JSON(label="Generated Script")

        with gr.Column():
            player = gr.Audio(label="Audio Output", autoplay=True)
            transcript = gr.Textbox(label="Transcript")
            btn_play = gr.Button("▶️ Play Next Line")
            q_input = gr.Textbox(label="Ask a Question")
            btn_interrupt = gr.Button("✋ Interrupt Podcast")

    # Bind events
    btn_gen.click(generate_script, [pdf_input, persona], [status, script_display])
    btn_play.click(play_next_chunk, [], [player, transcript])
    btn_interrupt.click(interrupt_and_ask, [q_input], [player, transcript])

# 8️⃣ Launch
if __name__ == "__main__":
    demo.launch()