Delete app.py

app.py

@@ -1,205 +0,0 @@
-import concurrent.futures as cf
-import glob
-import io
-import os
-import time
-from pathlib import Path
-from tempfile import NamedTemporaryFile
-from typing import List, Literal
-import re
-import gradio as gr
-import PyPDF2  # <-- Make sure this is included
-from transformers import pipeline
-from pydantic import BaseModel
-
-# Initialize Hugging Face text generation model
-text_generator = pipeline('text-generation', model='EleutherAI/gpt-neo-2.7B')
-summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
-
-# Instruction templates
-INSTRUCTION_TEMPLATES = {
-    "podcast": {
-        "intro": """Your task is to take the input text provided and turn it into a lively, engaging, informative podcast dialogue, in the style of NPR...""",
-        "text_instructions": "First, carefully read through the input text...",
-        "scratch_pad": """Brainstorm creative ways to discuss the main topics...""",
-        "prelude": """Now that you have brainstormed ideas and created a rough outline...""",
-        "dialog": """Write a very long, engaging, informative podcast dialogue..."""
-    }
-}
-
-# Function to update instruction fields based on template selection
-def update_instructions(template):
-    return (
-        INSTRUCTION_TEMPLATES[template]["intro"],
-        INSTRUCTION_TEMPLATES[template]["text_instructions"],
-        INSTRUCTION_TEMPLATES[template]["scratch_pad"],
-        INSTRUCTION_TEMPLATES[template]["prelude"],
-        INSTRUCTION_TEMPLATES[template]["dialog"]
-    )
-
-# Define the structure of dialogue
-class DialogueItem(BaseModel):
-    text: str
-    speaker: Literal["speaker-1", "speaker-2"]
-
-class Dialogue(BaseModel):
-    scratchpad: str
-    dialogue: List[DialogueItem]
-
-# Function to read README.md
-def read_readme():
-    readme_path = Path("README.md")
-    if readme_path.exists():
-        with open(readme_path, "r") as file:
-            content = file.read()
-            content = re.sub(r'--.*?--', '', content, flags=re.DOTALL)
-            return content
-    else:
-        return "README.md not found. Please check the repository for more information."
-
-# Hugging Face-based dialogue generation function
-def generate_dialogue(text: str, intro_instructions: str, text_instructions: str, 
-                      scratch_pad_instructions: str, prelude_dialog: str, 
-                      podcast_dialog_instructions: str, edited_transcript: str = None, 
-                      user_feedback: str = None) -> str:
-    # Combine instructions and text into a prompt
-    full_prompt = f"""
-    {intro_instructions}
-    
-    Original text:
-    {text}
-    
-    {text_instructions}
-    
-    Brainstorming:
-    {scratch_pad_instructions}
-    
-    Prelude:
-    {prelude_dialog}
-    
-    Dialogue:
-    {podcast_dialog_instructions}
-    
-    {edited_transcript if edited_transcript else ""}
-    {user_feedback if user_feedback else ""}
-    """
-    
-    # Generate text using Hugging Face model
-    generated = text_generator(full_prompt, max_length=1000)  # Adjust max_length as needed
-    return generated[0]['generated_text']  # Extract generated text from the response
-
-# Function to handle audio generation (could be expanded later)
-def get_mp3(text: str, voice: str, audio_model: str) -> bytes:
-    # Placeholder for audio generation; currently not implemented
-    # You can use text-to-speech services or local TTS engines
-    return b""
-
-# Main audio generation function (adapted for Hugging Face text generation)
-def generate_audio(
-    file: str,
-    text_model: str = "EleutherAI/gpt-neo-2.7B",
-    audio_model: str = "tts-1",
-    speaker_1_voice: str = "alloy",
-    speaker_2_voice: str = "echo",
-    intro_instructions: str = '',
-    text_instructions: str = '',
-    scratch_pad_instructions: str = '',
-    prelude_dialog: str = '',
-    podcast_dialog_instructions: str = '',
-    edited_transcript: str = None,
-    user_feedback: str = None,
-    original_text: str = None,
-    debug = False,
-) -> tuple:
-    
-    # Combine input text from the single file
-    combined_text = original_text or ""
-    if not combined_text:
-        with Path(file).open("rb") as f:
-            text = f.read().decode('utf-8')  # Assuming the PDF text is extracted as UTF-8
-            combined_text += text + "\n\n"
-
-    # Generate the dialogue using Hugging Face
-    llm_output = generate_dialogue(
-        combined_text,
-        intro_instructions=intro_instructions,
-        text_instructions=text_instructions,
-        scratch_pad_instructions=scratch_pad_instructions,
-        prelude_dialog=prelude_dialog,
-        podcast_dialog_instructions=podcast_dialog_instructions,
-        edited_transcript=edited_transcript,
-        user_feedback=user_feedback
-    )
-
-    # Placeholder for audio (since TTS implementation is omitted)
-    audio = b""
-    transcript = llm_output
-    characters = len(llm_output)
-
-    # Generating audio (placeholder logic)
-    with cf.ThreadPoolExecutor() as executor:
-        futures = []
-        for line in llm_output.split('\n'):
-            future = executor.submit(get_mp3, line, speaker_1_voice, audio_model)
-            futures.append(future)
-            characters += len(line)
-
-        for future in futures:
-            audio_chunk = future.result()
-            audio += audio_chunk
-
-    temporary_directory = "./tmp/"
-    os.makedirs(temporary_directory, exist_ok=True)
-
-    # Save audio to a temporary file
-    temporary_file = NamedTemporaryFile(dir=temporary_directory, delete=False, suffix=".mp3")
-    temporary_file.write(audio)
-    temporary_file.close()
-
-    return temporary_file.name, transcript, combined_text
-
-# Gradio Interface (for single file upload)
-def generate_podcast_interface(file, intro_instructions, text_instructions, scratch_pad_instructions, prelude_dialog, podcast_dialog_instructions):
-    # Handle a single PDF file
-    combined_text = ""
-    file_path = file.name
-    with open(file_path, "rb") as f:
-        reader = PyPDF2.PdfReader(f)
-        for page in reader.pages:
-            combined_text += page.extract_text() + "\n\n"
-    
-    # Generate podcast dialogue
-    podcast_dialogue = generate_dialogue(combined_text, intro_instructions, text_instructions, scratch_pad_instructions, prelude_dialog, podcast_dialog_instructions)
-    
-    return podcast_dialogue
-
-def interface():
-    with gr.Blocks() as demo:
-        gr.Markdown("# Podcast Generator from PDF File")
-        gr.Markdown("Upload a PDF file, input instructions, and generate podcast dialogues using Hugging Face models.")
-
-        with gr.Row():
-            pdf_input = gr.File(label="Upload PDF", file_types=[".pdf"])
-
-        with gr.Row():
-            intro_input = gr.Textbox(label="Intro Instructions", value=INSTRUCTION_TEMPLATES["podcast"]["intro"], lines=5)
-            text_instructions_input = gr.Textbox(label="Text Instructions", value=INSTRUCTION_TEMPLATES["podcast"]["text_instructions"], lines=5)
-            scratch_pad_input = gr.Textbox(label="Scratch Pad Instructions", value=INSTRUCTION_TEMPLATES["podcast"]["scratch_pad"], lines=5)
-            prelude_input = gr.Textbox(label="Prelude", value=INSTRUCTION_TEMPLATES["podcast"]["prelude"], lines=5)
-            dialog_input = gr.Textbox(label="Podcast Dialogue Instructions", value=INSTRUCTION_TEMPLATES["podcast"]["dialog"], lines=5)
-
-        generate_btn = gr.Button("Generate Podcast Dialogue")
-
-        output_text = gr.Textbox(label="Generated Podcast Dialogue", lines=10)
-
-        generate_btn.click(
-            fn=generate_podcast_interface,
-            inputs=[pdf_input, intro_input, text_instructions_input, scratch_pad_input, prelude_input, dialog_input],
-            outputs=output_text
-        )
-
-    return demo
-
-if __name__ == "__main__":
-    demo = interface()
-    demo.launch()