app.py

import gradio as gr
import os
import json
import requests
import fitz

# Streaming endpoint
API_URL = "https://api.openai.com/v1/chat/completions"  # os.getenv("API_URL") + "/generate_stream"

key = os.environ.get('key')

# Function to extract text from a PDF file
def extract_text_from_pdf(file_path):
    text = ""
    with fitz.open(file_path) as doc:
        for page in doc:
            text += page.get_text()
    return text


def extract_text_from_multiple_pdfs(files):
    all_text = ""
    for file in files:
        text = extract_text_from_pdf(file)
        all_text += text + "\n"  # Adding a newline for separation between documents
    return all_text


def predict(inputs, file_upload_1, file_upload_2, file_text_1, file_text_2, top_p, temperature, chat_counter,
            chatbot=[], history=[]):

    # Extract text from PDF files if provided
    if file_upload_1 is not None:
        file_uploaded_1_text = extract_text_from_multiple_pdfs(file_upload_1)
    else:
        file_uploaded_1_text = ''

    if file_upload_2 is not None:
        file_uploaded_2_text = extract_text_from_multiple_pdfs(file_upload_2)
    else:
        file_uploaded_2_text = ''

    file_text_1 = file_text_1 if file_text_1 is not None else ''
    file_text_2 = file_text_2 if file_text_2 is not None else ''

    prompta = 'You are SyncChain, an AI assistant who is an expert in discovering and describing collaboration ' \
              'opportunities between individuals. Make sure to follow any additional instructions at the end of any ' \
              'provided information. Instead of directly calling the name of the user, you will address the user as ' \
              'if you are talking to them. Make sure to use bold fonts for writing \'collaboration with x \' for each ' \
              'person, switch rows after you finish talking about each potential collaborator'
    promptb = 'Below are the information of the user who is interested in seeking collaboration'
    promptc = 'Below are the information of one or multiple people that might be of interest to the user, if there is ' \
              'information of only one person, analyze in detail how that person can collaborate with the user. If ' \
              'there are information about multiple people, then describe collaboration between the user and each of ' \
              'them, and provide a ranking for the multiple potential collaborators.'
    # Combine user inputs and extracted file texts
    combined_input = prompta + "\n" + promptb + "\n" + file_text_1 + "\n" + file_uploaded_1_text + "\n" + promptc + "\n" + file_text_2 + "\n" + file_uploaded_2_text + "\n" + inputs

    # Prepare the API request headers
    headers = {
        "Content-Type": "application/json",
        "Authorization": f"Bearer {key}"
    }

    # Construct the API payload
    payload = {
        "model": "gpt-4-turbo-preview",
        "messages": [{"role": "user", "content": combined_input}],
        "temperature": temperature,
        "top_p": top_p,
        "n": 1,
        "stream": False,
        "presence_penalty": 0,
        "frequency_penalty": 0,
    }

    # Send the request to the API and get the response
    api_response = requests.post(API_URL, headers=headers, json=payload)
    response_data = api_response.json()

    # Extract the assistant's response from the API response
    assistant_response = response_data.get('choices', [{}])[0].get('message', {}).get('content',
                                                                                      "I couldn't understand that.")

    # Update history and chat with the new interaction
    history.append(inputs)  # Append the combined user input to the history
    history.append(assistant_response)  # Append the assistant's response to the history
    chat = [(history[i], history[i + 1]) for i in
            range(0, len(history) - 1, 2)]  # Pair each user input with the corresponding assistant response

    # Append only the assistant's response to the chatbot component (user inputs are not displayed)
    chatbot.append({"role": "assistant", "content": assistant_response})

    # Increment the chat counter
    chat_counter += 1

    # Return the updated variables
    return chat, history, chat_counter


# Resetting to blank
def reset_textbox():
    return gr.update(value='')


# to set a component as visible=False
def set_visible_false():
    return gr.update(visible=False)


# to set a component as visible=True
def set_visible_true():
    return gr.update(visible=True)


title = """<h1 align="center">SyncChain Demo</h1>"""
# display message for themes feature
theme_addon_msg = """<center>🌟 This Demo also introduces you to Gradio Themes. Discover more on Gradio website using our <a href="https://gradio.app/theming-guide/" target="_blank">Themeing-Guide🎨</a>! You can develop from scratch, modify an existing Gradio theme, and share your themes with community by uploading them to huggingface-hub easily using <code>theme.push_to_hub()</code>.</center>
"""

# Using info to add additional information about System message in GPT4
system_msg_info = """A conversation could begin with a system message to gently instruct the assistant. 
System message helps set the behavior of the AI Assistant. For example, the assistant could be instructed with 'You are a helpful assistant.'"""

# Modifying existing Gradio Theme
theme = gr.themes.Soft(text_size=gr.themes.sizes.text_lg)
CSS = """
 footer{display:none !important}
.contain { display: flex; flex-direction: column; }
.gradio-container { height: 100vh !important; }
#component-0 { height: 100%; }
#chatbot { flex-grow: 1; overflow: auto;}
"""

with gr.Blocks(css=CSS,
               theme=theme) as demo:
    gr.HTML(title)
    with gr.Column(elem_id="col_container"):
        # Users need to provide their own GPT4 API key, it is no longer provided by Huggingface
        chatbot = gr.Chatbot(label='SyncChain AI', elem_id="chatbot",value=[[None, "This is SyncChain AI, feel free "
                                                                                   "to provide me with information of"
                                                                                   " yourself and any other people, "
                                                                                   "I'll tell you how you can "
                                                                                   "collaborate with them"]])
        inputs = gr.Textbox(placeholder="I want to know...", label="Enter any additional instruction or query here")
        state = gr.State([])
        with gr.Row():
            with gr.Column(scale=7):
                # b1 = gr.Button().style(full_width=True)
                b1 = gr.Button()

        # top_p, temperature
        top_p = gr.Number(value=1, visible=False, precision=0)
        temperature = gr.Number(value=1, visible=False, precision=0)
        chat_counter = gr.Number(value=0, visible=False, precision=0)

        # Add a file upload component
        with gr.Row():
            file_text_1 = gr.Textbox(placeholder="Enter here",
                                     label="Information about yourself, you can also upload files below")
            file_text_2 = gr.Textbox(placeholder="Enter here",
                                     label="Information about other people of interest, you can also upload files below")
        with gr.Row():
            file_upload_1 = gr.File(label="Upload PDF documents here", file_types=["pdf"], file_count='multiple')
            file_upload_2 = gr.File(label="Upload PDF documents here", file_types=["pdf"], file_count='multiple')

    # Event handling
    inputs.submit(predict,
                  [inputs, file_upload_1, file_upload_2, file_text_1, file_text_2, top_p, temperature, chat_counter,
                   chatbot, state], [chatbot, state, chat_counter], )  # openai_api_key
    b1.click(predict,
             [inputs, file_upload_1, file_upload_2, file_text_1, file_text_2, top_p, temperature, chat_counter, chatbot,
              state], [chatbot, state, chat_counter], )  # openai_api_key
    b1.click(reset_textbox, [], [inputs])
    inputs.submit(reset_textbox, [], [inputs])

# demo.queue(max_size=99, concurrency_count=20).launch(debug=True)
demo.queue(max_size=99).launch(debug=True)