app.py

import gradio as gr
from langchain.chains import LLMChain
from langchain.prompts import PromptTemplate
from langchain.document_loaders import WebBaseLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.llms import HuggingFacePipeline
from transformers import pipeline
import tempfile
import os

# Step 1: CPU-friendly summarization LLM (Flan-T5 Small)
summary_pipe = pipeline("text2text-generation", model="google/flan-t5-base", device=-1)
llm = HuggingFacePipeline(pipeline=summary_pipe)

# Step 2: Summarization Prompt
summary_prompt = PromptTemplate.from_template("""
Summarize the following webpage content in a clear, concise way:

{text}

Summary:
""")

summary_chain = LLMChain(llm=llm, prompt=summary_prompt)

# Step 3: URL to Text -> Summarize -> Text to Speech

def url_to_audio_summary(url):
    try:
        # Load and split text
        loader = WebBaseLoader(url)
        docs = loader.load()
        splitter = RecursiveCharacterTextSplitter(chunk_size=2000, chunk_overlap=100)
        splits = splitter.split_documents(docs)
        
        full_text = "\n".join([s.page_content for s in splits])
        
        # Summarize
        summary = summary_chain.run(text=full_text)

        # Text to Speech
        tts_pipe = pipeline("text-to-speech", model="espnet/kan-bayashi_ljspeech_vits", device=-1)
        audio = tts_pipe(summary)["audio"]

        # Save audio to temp WAV
        with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as f:
            f.write(audio)
            audio_path = f.name

        return summary, audio_path

    except Exception as e:
        return f"Error: {str(e)}", None

# Step 4: Gradio Interface
iface = gr.Interface(
    fn=url_to_audio_summary,
    inputs=gr.Textbox(label="Article URL", placeholder="Paste a news/blog URL here..."),
    outputs=[
        gr.Textbox(label="Summary"),
        gr.Audio(label="Audio Summary")
    ],
    title="🗣️ URL to Audio Summary Agent",
    description="An agent that reads web articles and gives you an audio summary. CPU-only. Built with LangChain + Hugging Face."
)

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