app.py

import gradio as gr
from huggingface_hub import InferenceClient

client = InferenceClient("Qwen/Qwen2.5-7B-Instruct")

def respond(message, history):
    
    messages = [{"role": "system", "content": "You are a gym bro that is passionate about meal prepping. Help people create fitness goals, create healthy life styles, and eat well. Responses should be under 600 words. Keep language simple and 'bro-y'.understand the user's goal before giving recommendations."}] 
        #content is the system message
    
    if history:
        messages.extend(history)
    
    messages.append({"role": "user", "content": message})

    response = ""  #empty string and indent client.chat
    
    stream = client.chat_completion(  #turn into a stream to arrive piece by piece
        messages,
        max_tokens=600,
        temperature=0.3, # Higher temperature values like 0.8 will make the output more random, 
            # while lower values like 0.2 will make it more focused and deterministic.
        stream=True  #pasted into client.chat
    )
    
    for message in stream: #add a for loop
        token = message.choices[0].delta.content #paste in
        response += token
        yield response #change to yield and get rid of extra parts that are now in for loop
    
# Customize colors, sizing and fonts
custom_theme = gr.themes.Soft(
    primary_hue="pink",
    secondary_hue="fuchsia", 
    neutral_hue="gray",
    spacing_size="lg",
    radius_size="lg",
    text_size="lg",
    font=[gr.themes.GoogleFont("IBM Plex Sans"), "sans-serif"],
    font_mono=[gr.themes.GoogleFont("IBM Plex Mono"), "monospace"]
)

chatbot = gr.ChatInterface(respond,  type="messages", theme=custom_theme)

chatbot.launch()