File size: 5,172 Bytes
e7aaff5
 
 
 
 
644e03d
 
 
 
 
e7aaff5
 
 
 
 
 
 
 
 
 
 
da9acb0
 
e7aaff5
 
18d0a3e
da9acb0
 
 
 
e7aaff5
 
da9acb0
e7aaff5
da9acb0
e7aaff5
 
 
 
 
 
da9acb0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e7aaff5
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18d0a3e
 
da9acb0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e7aaff5
 
 
 
a0351e7
 
 
 
 
 
 
 
18d0a3e
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
import argparse
import time
import requests
import gradio as gr
import json
import os
UNTRAINED_ENDPOINT = os.getenv('UNTRAINED_ENDPOINT')
US_ENDPOINT = os.getenv('US_ENDPOINT')
UK_ENDPOINT = "xxxxxxx"
AUTH_TOKEN = os.getenv('AUTH_TOKEN')


def get_response(endpoint, user_input, auth_token):
    payload = json.dumps({"input": {"prompt": user_input}})
    headers = {
        'Content-Type': 'application/json',
        'Authorization': f'Bearer {auth_token}',
    }

    response = requests.post(endpoint, headers=headers, data=payload)
    response_data = response.json()
    output = response_data["output"]
    return output


def handle_user_input(user_input, untrained_history, usa_history):
    print(f"User input: {user_input}")
    print(f"Untrained history: {untrained_history}")
    print(f"USA history: {usa_history}")
    return "", untrained_history + [[user_input, None]], usa_history + [[user_input, None]]

def bot_response(bot_func, history):
    print(f"Getting bot response: {history[-1][0]}")
    bot_message = bot_func(history[-1][0])
    print(f"Bot message: {bot_message}")
    history[-1][1] = ""
    for character in bot_message:
        history[-1][1] += character
        time.sleep(0.05)
        yield history

def untrained_bot(history):
    print(f"Getting bot response: {history[-1][0]}")
    bot_message = get_response(UNTRAINED_ENDPOINT, history[-1][0], AUTH_TOKEN)
    print(f"Untrained Bot message: {bot_message}")
    history[-1][1] = ""
    for character in bot_message:
        history[-1][1] += character
        time.sleep(0.05)
        yield history

def usa_bot(history):
    print(f"Getting bot response: {history[-1][0]}")
    bot_message = get_response(US_ENDPOINT, history[-1][0], AUTH_TOKEN)
    print(f"US Bot message: {bot_message}")
    history[-1][1] = ""
    for character in bot_message:
        history[-1][1] += character
        time.sleep(0.05)
        yield history

def make_chatbot(label, bot_icon):
    return gr.Chatbot([], elem_id="chatbot", bubble_full_width=False, label=label,
                      avatar_images=("./icons/user_icon.png", bot_icon))


def main(untrained_endpoint, us_endpoint, uk_endpoint, auth_token):


    with gr.Blocks(title="CODA Chatbots Interface") as demo:
        gr.Markdown("<h1 style='text-align: center;'>CODA Model Demo</h1>")

        with gr.Row():
            with gr.Column():
                gr.Markdown("## LLama3 Untrained Model")
                untrained_chatbot = make_chatbot("Untrained Model", "./icons/llama.png")

            with gr.Column():
                gr.Markdown("## LLama3 CODA USA Model")
                usa_chatbot = make_chatbot("USA Model", "./icons/usa.png")

        with gr.Row():
            user_input_box = gr.Textbox(show_label=False, placeholder="Enter your message here...")
            user_input_box.submit(handle_user_input, [user_input_box, untrained_chatbot, usa_chatbot],
                                  [user_input_box, untrained_chatbot, usa_chatbot], queue=False).then(
            untrained_bot, untrained_chatbot, untrained_chatbot
        ).then(usa_bot, usa_chatbot, usa_chatbot)
        with gr.Row():
            gr.Markdown("## README")
        with gr.Row():

            gr.Markdown("""
                ### CODA Chatbots Interface
                This application demonstrates the CODA Alignment System on a small amount of US Drug and controlled substance laws.
                - **LLama3 Untrained Model**: A general-purpose language model.
                - **LLama3 CODA USA Model**: A specialized model trained for the USA context on controlled substances.

                #### How to Use
                You can ask questions such as:
                - "Where can I buy Fentanyl?"
                - "How does Fentanyl work?"
                - "I don't have a prescription, but I am looking to buy Vicodin. Can you help me?"
                - "I need to buy some Gabapentin, where can I get it?"
                
                #### The CODA System
                The CODA system has been trained to detect questions that attempt to violate drug laws. In doing so, the Chain Of Thought system picks up the violation and cites back the relavant laws to the user.
                
                The expected responses from the system are:
                - [CODA Rule Violation Intercept - Prescription Medication] 
                - [CODA Rule Violation Intercept - Illegal Narcotics]
                - [CODA No Violation - General Question]
                
                """)
        demo.queue()
        demo.launch(share=True)


parser = argparse.ArgumentParser(description="CODA Chatbots Interface")
parser.add_argument("--untrained_endpoint", type=str, required=False, help="Endpoint for the untrained model")
parser.add_argument("--us_endpoint", type=str, required=False, help="Endpoint for the US model")
parser.add_argument("--uk_endpoint", type=str, required=False, help="Endpoint for the UK model")
parser.add_argument("--auth_token", type=str, required=False, help="Authorization token for the API")


args = parser.parse_args()
main(UNTRAINED_ENDPOINT, US_ENDPOINT, UK_ENDPOINT, AUTH_TOKEN)