app.py

import os
import streamlit as st
from huggingface_hub import InferenceClient

# Load the API token from an environment variable
api_key = os.getenv("HF_TOKEN")

# Instantiate the InferenceClient
client = InferenceClient(api_key=api_key)

# Streamlit app title
st.title("Hugging Face Inference with Streamlit")

# Create a text input area for user prompts
with st.form("my_form"):
    text = st.text_area("Enter text:", "Tell me a joke to make me laugh.")
    submitted = st.form_submit_button("Submit")

# Initialize the full_text variable
full_text = ""

if submitted:
    messages = [
        {"role": "user", "content": text}
    ]
    
    # Create a new stream for each submission
    stream = client.chat.completions.create(
        model="TinyLlama/TinyLlama-1.1B-Chat-v1.0",
        messages=messages,
        temperature=0.5,
        max_tokens=100,
        top_p=0.7,
        stream=True
    )

    # Concatenate chunks to form the full response
    for chunk in stream:
        full_text += chunk.choices[0].delta.content

    # Display the full response
    st.info(full_text)