|
import streamlit as st |
|
from openai import OpenAI, APIError |
|
import os |
|
import tenacity |
|
|
|
|
|
ACCESS_TOKEN = os.getenv("HF_TOKEN") |
|
|
|
|
|
MODEL = "NousResearch/Hermes-3-Llama-3.1-8B" |
|
|
|
client = OpenAI( |
|
base_url="https://api-inference.huggingface.co/v1/", |
|
api_key=ACCESS_TOKEN, |
|
) |
|
|
|
@tenacity.retry(wait=tenacity.wait_exponential(multiplier=1, min=4, max=10)) |
|
def respond( |
|
message, |
|
history: list[tuple[str, str]], |
|
system_message, |
|
max_tokens, |
|
temperature, |
|
top_p, |
|
): |
|
try: |
|
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.completions.create( |
|
model=MODEL, |
|
max_tokens=max_tokens, |
|
stream=True, |
|
temperature=temperature, |
|
top_p=top_p, |
|
messages=messages, |
|
): |
|
token = message.choices[0].delta.content |
|
|
|
response += token |
|
yield response |
|
except APIError as e: |
|
error_details = e.body |
|
error_type = error_details.get("type") |
|
error_code = error_details.get("code") |
|
error_param = error_details.get("param") |
|
error_message = error_details.get("message") |
|
|
|
if error_type: |
|
error_str = f"{error_type}: {error_message} (code: {error_code}, param: {error_param})" |
|
else: |
|
error_str = "An error occurred during streaming" |
|
|
|
print(f"Error: {error_str}") |
|
yield error_str |
|
except Exception as e: |
|
print(f"Error: {e}") |
|
yield "Error occurred. Please try again." |
|
|
|
st.title("Chatbot") |
|
st.write("Type a message to chat with the bot") |
|
|
|
message = st.text_input("Message") |
|
history = [] |
|
system_message = st.text_input("System message") |
|
max_tokens = st.slider("Max new tokens", 1, 2048, 2048) |
|
temperature = st.slider("Temperature", 0.1, 4.0, 0.7) |
|
top_p = st.slider("Top-P", 0.1, 1.0, 0.95) |
|
|
|
if st.button("Send"): |
|
response = respond(message, history, system_message, max_tokens, temperature, top_p) |
|
for resp in response: |
|
st.write(resp) |