File size: 8,227 Bytes
6116c39
 
 
 
 
 
2394e8b
6116c39
 
 
 
 
 
 
 
e1f6aec
 
6116c39
 
 
 
 
e1f6aec
6116c39
e1f6aec
 
6116c39
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e1f6aec
6116c39
 
 
 
 
e1f6aec
 
 
 
 
6116c39
 
e1f6aec
 
 
 
 
6116c39
 
 
 
 
 
 
e1f6aec
6116c39
 
 
e1f6aec
 
6116c39
e1f6aec
 
 
6116c39
e1f6aec
6116c39
e1f6aec
 
6116c39
e1f6aec
 
 
 
 
 
 
 
 
 
 
 
 
 
6116c39
e1f6aec
 
6116c39
e1f6aec
 
 
 
 
 
6116c39
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
import os
import requests
import time
import streamlit as st

# Get the Hugging Face API Token from environment variables
HF_API_TOKEN = os.getenv("HF_API_KEY")
if not HF_API_TOKEN:
    raise ValueError("Hugging Face API Token is not set in the environment variables.")

# Hugging Face API URLs and headers for models
MISTRAL_API_URL = "https://api-inference.huggingface.co/models/mistralai/Mixtral-8x7B-Instruct-v0.1"
MINICHAT_API_URL = "https://api-inference.huggingface.co/models/GeneZC/MiniChat-2-3B"
DIALOGPT_API_URL = "https://api-inference.huggingface.co/models/microsoft/DialoGPT-large"
PHI3_API_URL = "https://api-inference.huggingface.co/models/microsoft/Phi-3-mini-4k-instruct"
GEMMA_API_URL = "https://api-inference.huggingface.co/models/google/gemma-1.1-7b-it"
GEMMA_2B_API_URL = "https://api-inference.huggingface.co/models/google/gemma-1.1-2b-it"
META_LLAMA_70B_API_URL = "https://api-inference.huggingface.co/models/meta-llama/Meta-Llama-3-70B-Instruct"
META_LLAMA_8B_API_URL = "https://api-inference.huggingface.co/models/meta-llama/Meta-Llama-3-8B-Instruct"
GEMMA_27B_API_URL = "https://api-inference.huggingface.co/models/google/gemma-2-27b"
GEMMA_27B_IT_API_URL = "https://api-inference.huggingface.co/models/google/gemma-2-27b-it"

HEADERS = {"Authorization": f"Bearer {HF_API_TOKEN}"}

def query_model(api_url, payload):
    response = requests.post(api_url, headers=HEADERS, json=payload)
    return response.json()

def count_tokens(text):
    return len(text.split())

MAX_TOKENS_PER_MINUTE = 1000
token_count = 0
start_time = time.time()

def handle_token_limit(text):
    global token_count, start_time
    current_time = time.time()
    if current_time - start_time > 60:
        token_count = 0
        start_time = current_time
    token_count += count_tokens(text)
    if token_count > MAX_TOKENS_PER_MINUTE:
        raise ValueError("Token limit exceeded. Please wait before sending more messages.")

def add_message_to_conversation(user_message, bot_message, model_name):
    st.session_state.conversation.append((user_message, bot_message, model_name))

# Streamlit app
st.set_page_config(page_title="Multi-LLM Chatbot Interface", layout="wide")
st.title("Multi-LLM Chatbot Interface")
st.write("Multi LLM-Chatbot Interface")

# Initialize session state for conversation and model history
if "conversation" not in st.session_state:
    st.session_state.conversation = []
if "model_history" not in st.session_state:
    st.session_state.model_history = {model: [] for model in [
        "Mistral-8x7B", "MiniChat-2-3B", "DialoGPT (GPT-2-1.5B)", "Phi-3-mini-4k-instruct",
        "Gemma-1.1-7B", "Gemma-1.1-2B", "Meta-Llama-3-70B-Instruct", "Meta-Llama-3-8B-Instruct",
        "Gemma-2-27B", "Gemma-2-27B-IT"
    ]}

# Dropdown for LLM selection
llm_selection = st.selectbox("Select Language Model", [
    "Mistral-8x7B", "MiniChat-2-3B", "DialoGPT (GPT-2-1.5B)", "Phi-3-mini-4k-instruct",
    "Gemma-1.1-7B", "Gemma-1.1-2B", "Meta-Llama-3-70B-Instruct", "Meta-Llama-3-8B-Instruct",
    "Gemma-2-27B", "Gemma-2-27B-IT"
])

# User input for question
question = st.text_input("Question", placeholder="Enter your question here...")

# Handle user input and LLM response
if st.button("Send") and question:
    try:
        handle_token_limit(question)  # Check token limit before processing
        with st.spinner("Waiting for the model to respond..."):
            chat_history = " ".join(st.session_state.model_history[llm_selection]) + f"User: {question}\n"
            if llm_selection == "Mistral-8x7B":
                response = query_model(MISTRAL_API_URL, {"inputs": chat_history})
                answer = response[0].get("generated_text", "No response") if isinstance(response, list) else "No response"
            elif llm_selection == "MiniChat-2-3B":
                response = query_model(MINICHAT_API_URL, {"inputs": chat_history})
                if "error" in response and "is currently loading" in response["error"]:
                    answer = f"Model is loading, please wait {response['estimated_time']} seconds."
                else:
                    answer = response[0].get("generated_text", "No response") if isinstance(response, list) else "No response"
            elif llm_selection == "DialoGPT (GPT-2-1.5B)":
                response = query_model(DIALOGPT_API_URL, {"inputs": chat_history})
                answer = response.get("generated_text", "No response") if isinstance(response, dict) else response[0].get("generated_text", "No response") if isinstance(response, list) else "No response"
            elif llm_selection == "Phi-3-mini-4k-instruct":
                response = query_model(PHI3_API_URL, {"inputs": chat_history})
                answer = response[0].get("generated_text", "No response") if isinstance(response, list) else "No response"
            elif llm_selection == "Gemma-1.1-7B":
                response = query_model(GEMMA_API_URL, {"inputs": chat_history})
                answer = response.get("generated_text", "No response") if isinstance(response, dict) else response[0].get("generated_text", "No response") if isinstance(response, list) else "No response"
            elif llm_selection == "Gemma-1.1-2B":
                response = query_model(GEMMA_2B_API_URL, {"inputs": chat_history})
                answer = response.get("generated_text", "No response") if isinstance(response, dict) else response[0].get("generated_text", "No response") if isinstance(response, list) else "No response"
            elif llm_selection == "Meta-Llama-3-70B-Instruct":
                response = query_model(META_LLAMA_70B_API_URL, {"inputs": chat_history})
                answer = response.get("generated_text", "No response") if isinstance(response, dict) else response[0].get("generated_text", "No response") if isinstance(response, list) else "No response"
            elif llm_selection == "Meta-Llama-3-8B-Instruct":
                response = query_model(META_LLAMA_8B_API_URL, {"inputs": chat_history})
                answer = response.get("generated_text", "No response") if isinstance(response, dict) else response[0].get("generated_text", "No response") if isinstance(response, list) else "No response"
            elif llm_selection == "Gemma-2-27B":
                response = query_model(GEMMA_27B_API_URL, {"inputs": chat_history})
                answer = response.get("generated_text", "No response") if isinstance(response, dict) else response[0].get("generated_text", "No response") if isinstance(response, list) else "No response"
            elif llm_selection == "Gemma-2-27B-IT":
                response = query_model(GEMMA_27B_IT_API_URL, {"inputs": chat_history})
                answer = response.get("generated_text", "No response") if isinstance(response, dict) else response[0].get("generated_text", "No response") if isinstance(response, list) else "No response"

            handle_token_limit(answer)  # Check token limit for output
            add_message_to_conversation(question, answer, llm_selection)
            st.session_state.model_history[llm_selection].append(f"User: {question}\n{llm_selection}: {answer}\n")
    except ValueError as e:
        st.error(str(e))

# Custom CSS for chat bubbles
st.markdown(
    """
    <style>
    .chat-bubble {
        padding: 10px 14px;
        border-radius: 14px;
        margin-bottom: 10px;
        display: inline-block;
        max-width: 80%;
        color: black;
    }
    .chat-bubble.user {
        background-color: #dcf8c6;
        align-self: flex-end;
    }
    .chat-bubble.bot {
        background-color: #fff;
        align-self: flex-start;
    }
    .chat-container {
        display: flex;
        flex-direction: column;
        gap: 10px;
        margin-top: 20px;
    }
    </style>
    """,
    unsafe_allow_html=True
)

# Display the conversation
st.write('<div class="chat-container">', unsafe_allow_html=True)
for user_message, bot_message, model_name in st.session_state.conversation:
    st.write(f'<div class="chat-bubble user">You: {user_message}</div>', unsafe_allow_html=True)
    st.write(f'<div class="chat-bubble bot">{model_name}: {bot_message}</div>', unsafe_allow_html=True)
st.write('</div>', unsafe_allow_html=True)