app.py

import os
os.system('pip install transformers')
os.system('pip install datasets')
os.system('pip install gradio')
os.system('pip install minijinja')

import gradio as gr
from huggingface_hub import InferenceClient
from transformers import pipeline
from datasets import load_dataset
import time

dataset = load_dataset("ibunescu/qa_legal_dataset_train")

# Use a pipeline as a high-level helper
pipe = pipeline("fill-mask", model="nlpaueb/legal-bert-base-uncased")

client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,
    max_tokens,
    temperature,
    top_p,
):
    messages = [{"role": "system", "content": system_message}]

    for val in history:
        if val[0]:
            messages.append({"role": "user", "content": val[0]})
        if val[1]:
            messages.append({"role": "assistant", "content": val[1]})

    messages.append({"role": "user", "content": message})

    response = ""
    for message in client.chat_completion(
        messages,
        max_tokens=max_tokens,
        stream=True,
        temperature=temperature,
        top_p=top_p,
    ):
        token = message.choices[0].delta.content
        if token is not None:
            response += token
        yield response, history + [(message, response)]

def score_argument(argument):
    merits_keywords = ["compelling", "convincing", "strong", "solid"]
    laws_keywords = ["statute", "law", "regulation", "act"]
    precedents_keywords = ["precedent", "case", "ruling", "decision"]
    verdict_keywords = ["guilty", "innocent", "verdict", "judgment"]

    merits_score = sum([1 for word in merits_keywords if word in argument.lower()])
    laws_score = sum([1 for word in laws_keywords if word in argument.lower()])
    precedents_score = sum([1 for word in precedents_keywords if word in argument.lower()])
    verdict_score = sum([1 for word in verdict_keywords if word in argument.lower()])
    length_score = len(argument.split())

    merits_value = merits_score * 2
    laws_value = laws_score * 3
    precedents_value = precedents_score * 4
    verdict_value = verdict_score * 5

    total_score = merits_value + laws_value + precedents_value + verdict_value + length_score

    return total_score

def color_code(score):
    if score > 50:
        return "green"
    elif score > 30:
        return "yellow"
    else:
        return "red"

# Custom CSS for white background and black text for input and output boxes
custom_css = """
body {
    background-color: #ffffff;
    color: #000000;
    font-family: Arial, sans-serif;
}
.gradio-container {
    max-width: 1000px;
    margin: 0 auto;
    padding: 20px;
    background-color: #ffffff;
    border: 1px solid #e0e0e0;
    border-radius: 8px;
    box-shadow: 0 2px 5px rgba(0, 0, 0, 0.1);
}
.gr-button {
    background-color: #ffffff !important;
    border-color: #ffffff !important;
    color: #000000 !important;
}
.gr-button:hover {
    background-color: #ffffff !important;
    border-color: #004085 !important;
}
.gr-input, .gr-textbox, .gr-slider, .gr-markdown, .gr-chatbox {
    border-radius: 4px;
    border: 1px solid #ced4da;
    background-color: #ffffff !important;
    color: #000000 !important;
}
.gr-input:focus, .gr-textbox:focus, .gr-slider:focus {
    border-color: #ffffff;
    outline: 0;
    box-shadow: 0 0 0 0.2rem rgba(255, 255, 255, 1.0);
}
#flagging-button {
    display: none;
}
footer {
    display: none;
}
.chatbox .chat-container .chat-message {
    background-color: #ffffff !important;
    color: #000000 !important;
}
.chatbox .chat-container .chat-message-input {
    background-color: #ffffff !important;
    color: #000000 !important;
}
.gr-markdown {
    background-color: #ffffff !important;
    color: #000000 !important;
}
.gr-markdown h1, .gr-markdown h2, .gr-markdown h3, .gr-markdown h4, .gr-markdown h5, .gr-markdown h6, .gr-markdown p, .gr-markdown ul, .gr-markdown ol, .gr-markdown li {
    color: #000000 !important;
}
.score-box {
    width: 60px;
    height: 60px;
    display: flex;
    align-items: center;
    justify-content: center;
    font-size: 12px;
    font-weight: bold;
    color: black;
    margin: 5px;
}
.scroll-box {
    max-height: 200px;
    overflow-y: scroll;
    border: 1px solid #ced4da;
    padding: 10px;
    border-radius: 4px;
}
"""

# Function to facilitate the conversation between the two chatbots
def chat_between_bots(system_message1, system_message2, max_tokens, temperature, top_p, history1, history2, shared_history, message):
    response1, history1 = list(respond(message, history1, system_message1, max_tokens, temperature, top_p))[-1]
    response2, history2 = list(respond(message, history2, system_message2, max_tokens, temperature, top_p))[-1]
    shared_history.append(f"Prosecutor: {response1}")
    shared_history.append(f"Defense Attorney: {response2}")
    
    max_length = max(len(response1), len(response2))
    response1 = response1[:max_length]
    response2 = response2[:max_length]
    
    score1 = score_argument(response1)
    score2 = score_argument(response2)
    
    prosecutor_color = color_code(score1)
    defense_color = color_code(score2)
    
    prosecutor_score_color = f"<div class='score-box' style='background-color:{prosecutor_color};'>Score: {score1}</div>"
    defense_score_color = f"<div class='score-box' style='background-color:{defense_color};'>Score: {score2}</div>"
    
    return response1, response2, history1, history2, shared_history, f"{response1}\n\n{response2}", prosecutor_score_color, defense_score_color

def update_pdf_gallery(pdf_files):
    return pdf_files

def add_message(history, message):
    for x in message["files"]:
        history.append(((x,), None))
    if message["text"] is not None:
        history.append((message["text"], None))
    return history, gr.MultimodalTextbox(value=None, interactive=False)

def bot(history):
    response = "**That's cool!**"
    history[-1][1] = ""
    for character in response:
        history[-1][1] += character
        time.sleep(0.05)
        yield history

def print_like_dislike(x: gr.LikeData):
    print(x.index, x.value, x.liked)

with gr.Blocks(css=custom_css) as demo:
    history1 = gr.State([])
    history2 = gr.State([])
    shared_history = gr.State([])
    pdf_files = gr.State([])
    
    with gr.Tab("Argument Evaluation"):
        message = gr.Textbox(label="Case to Argue")
        system_message1 = gr.State("You are an expert Prosecutor. Give your best arguments for the case on behalf of the prosecution.")
        system_message2 = gr.State("You are an expert Defense Attorney. Give your best arguments for the case on behalf of the Defense.")
        max_tokens = gr.State(512)
        temperature = gr.State(0.6)
        top_p = gr.State(0.95)
        
        with gr.Row():
            with gr.Column(scale=4):
                prosecutor_response = gr.Textbox(label="Prosecutor's Response", interactive=True)
                prosecutor_response.style(container=True).add_class("scroll-box")
            with gr.Column(scale=1):
                prosecutor_score_color = gr.HTML()
            
            with gr.Column(scale=4):
                defense_response = gr.Textbox(label="Defense Attorney's Response", interactive=True)
                defense_response.style(container=True).add_class("scroll-box")
            with gr.Column(scale=1):
                defense_score_color = gr.HTML()
        
        shared_argument = gr.Textbox(label="Case Outcome", interactive=True)
        submit_btn = gr.Button("Argue")
        
        submit_btn.click(chat_between_bots, inputs=[system_message1, system_message2, max_tokens, temperature, top_p, history1, history2, shared_history, message], outputs=[prosecutor_response, defense_response, history1, history2, shared_history, shared_argument, prosecutor_score_color, defense_score_color])
    
    with gr.Tab("PDF Management"):
        pdf_upload = gr.File(label="Upload Case Files (PDF)", file_types=[".pdf"])
        pdf_gallery = gr.Gallery(label="PDF Gallery")
        pdf_upload_btn = gr.Button("Update PDF Gallery")
        
        pdf_upload_btn.click(update_pdf_gallery, inputs=[pdf_upload], outputs=[pdf_gallery, pdf_files])
    
    with gr.Tab("Chatbot"):
        chatbot = gr.Chatbot(
            [],
            elem_id="chatbot",
            bubble_full_width=False
        )

        chat_input = gr.MultimodalTextbox(interactive=True, file_types=["image"], placeholder="Enter message or upload file...", show_label=False)

        chat_msg = chat_input.submit(add_message, [chatbot, chat_input], [chatbot, chat_input])
        bot_msg = chat_msg.then(bot, chatbot, chatbot, api_name="bot_response")
        bot_msg.then(lambda: gr.MultimodalTextbox(interactive=True), None, [chat_input])

        chatbot.like(print_like_dislike, None, None)

demo.queue()
demo.launch()