app.py

import gradio as gr
import json
import markdown
from telegraph import Telegraph
from gradio_client import Client
import time

# Set up the Telegraph client
telegraph = Telegraph()
telegraph.create_account(short_name='BookMindAI')

with open('detail_queries.json', 'r') as file:
    detail_queries = json.load(file)
with open('lang.json', 'r') as file:
    languages = [str(x) for x in json.load(file).keys()]

def markdown_to_html(md_content):
    return markdown.markdown(md_content)



def predict(input, images = []):
  client = Client("https://roboflow-gemini.hf.space/--replicas/bkd57/")
  result = client.predict(
      None,	
      images,	
      0.4,	
      2048,	
      "",	
      32,	
      1,	
      [[input,None]],
                api_name="/bot"
  )
  return result[0][1]
    
def fetch_summary(book_name, author, language):
    question = f"Provide a short summary of the book '{book_name}' by {author} in {language} language."
    answer = predict(question)
    return answer

def post_to_telegraph(title, content):
    html_content = markdown_to_html(content)
    response = telegraph.create_page(
        title=title,
        html_content=html_content
    )
    return 'https://telegra.ph/{}'.format(response['path'])

def generate_predictions(book_name, author, language_choice, detail_options=[]):

    details = ""
    for option in detail_options:
        query_template = detail_queries.get(option).format(book_name=book_name, author=author) + '. Answer in ' + language_choice[3:]
        try:
            response = predict(query_template)
            details += f"\n\n**{option}**:\n{response}"
        except:
            time.sleep(2)
            try:
                response = predict(query_template)
                details += f"\n\n**{option}**:\n{response}"
            except:
                pass
            
    summary = fetch_summary(book_name, author, language_choice[3:])
    combined_summary = summary + details
    try:
        telegraph_url = post_to_telegraph(f"Summary of {book_name} by {author}", combined_summary)
    except requests.exceptions.ConnectionError:
        telegraph_url = "Error connecting to Telegraph API"

    return combined_summary, telegraph_url

with gr.Blocks(title="📚 BookMindAI", theme=gr.themes.Base()).queue() as demo:
    gr.DuplicateButton()
    with gr.Tab("Summarize book🎯"):
        with gr.Row():
            with gr.Column():
                book_name_input = gr.Textbox(placeholder="Enter Book Name", label="Book Name")
                author_name_input = gr.Textbox(placeholder="Enter Author Name", label="Author Name")
                language_input = gr.Dropdown(choices=languages, label="Language")
                detail_options_input = gr.CheckboxGroup(choices=list(detail_queries.keys()), label="Details to Include", visible=True)
                run_button_summarize = gr.Button("Run", visible=True)

            with gr.Column():
                telegraph_link_output = gr.Markdown(label="View on Telegraph", visible=True)
        with gr.Row():
            summary_output = gr.Markdown(label="Parsed Content", visible=True)
        
        run_button_summarize.click(fn=generate_predictions,
                                   inputs=[book_name_input, author_name_input, language_input, detail_options_input],
                                   outputs=[summary_output, telegraph_link_output],
                                   show_progress=True, queue=True)

        examples_summarize = [
            ["Harry Potter and the Philosopher's Stone", "J.K. Rowling", "🇬🇧 english"],
            ["Pride and Prejudice", "Jane Austen", "🇺🇦 ukrainian"],
            ["The Great Gatsby", "F. Scott Fitzgerald", "🇫🇷 french"]
        ]
        gr.Examples(examples=examples_summarize, inputs=[book_name_input, author_name_input, language_input, detail_options_input])

    with gr.Tab("Talk about book🎓"):
        chat_examples = [
            "How do the underlying themes of a book reflect the societal values and beliefs of its time?",
            "In what ways do the characters' personal journeys mirror the broader human experience?"
        ]

        def chat_response(message, history):
            for i in range(len(message)):
                response = predict(message)
                yield response

        chat_interface = gr.ChatInterface(chat_response, examples=chat_examples, title='Talk with Gemini PRO about any book.')

demo.launch()