app.py

# Import the Gradio library for creating the web interface
import gradio as gr
# Import the InferenceClient from huggingface_hub to interact with the language model
from huggingface_hub import InferenceClient

# --- Configuration Constants ---
# Define the maximum number of tokens the model should generate in a single response
FIXED_MAX_TOKENS = 99999 # Note: This is a very high value, typical values are much lower (e.g., 512, 1024, 2048, 4096 for many models)


# --- Initialize the InferenceClient ---
# For custom OpenAI-compatible APIs, initialize the InferenceClient with the base URL.
# The specific model will be specified in the API call (e.g., chat_completion).
API_BASE_URL = "https://gtjr14qdt3vjwgdj.us-east-1.aws.endpoints.huggingface.cloud" # Base URL for the custom API

try:
    # Initialize the client with the base URL of your API.
    # If your API requires an authentication token, you might need to pass it here,
    # e.g., client = InferenceClient(base_url=API_BASE_URL, token="YOUR_API_TOKEN")
    # or ensure it's set as an environment variable if the client/API supports that.
    client = InferenceClient(base_url=API_BASE_URL)
    print(f"InferenceClient initialized with base_url: {API_BASE_URL}")
except Exception as e:
    print(f"Error initializing InferenceClient with base_url '{API_BASE_URL}': {e}")
    # Handle the error appropriately, e.g., by exiting or using a fallback
    raise RuntimeError(
        "Could not initialize InferenceClient. "
        f"Please check the API base URL ('{API_BASE_URL}') and ensure the server is accessible. "
        f"Error: {e}"
    )


# --- Core Chatbot Logic ---
def respond(message, history):
    """
    This function processes the user's message and the chat history to generate a response
    from the language model using the custom API.

    Args:
        message (str): The latest message from the user.
        history (list of lists): A list where each inner list contains a pair of
                                 [user_message, ai_message].

    Yields:
        str: The generated response token by token (for streaming).
    """
    # Initialize the messages list
    messages = []

    # Append past interactions from the history to the messages list
    # This provides context to the language model
    for user_message, ai_message in history:
        if user_message: # Ensure there's a user message
            messages.append({"role": "user", "content": user_message})
        if ai_message: # Ensure there's an AI message
            messages.append({"role": "assistant", "content": ai_message})

    # Append the current user's message to the messages list
    messages.append({"role": "user", "content": message})

    # Initialize an empty string to accumulate the response
    response_text = ""

    try:
        # Make a streaming call to the language model's chat completions endpoint.
        # The `model` parameter specifies which model to use at the endpoint.
        stream = client.chat_completion(
            messages=messages,              # The conversation history and current message
            max_tokens=FIXED_MAX_TOKENS,    # Maximum tokens for the response
            stream=True,                    # Enable streaming for token-by-token output
        )

        for chunk in stream:
            # Check if the chunk contains content and the content is not None
            # The exact structure of the chunk can vary based on the model/endpoint
            if chunk.choices and chunk.choices[0].delta and chunk.choices[0].delta.content is not None:
                token = chunk.choices[0].delta.content # Extract the token from the chunk
                response_text += token                 # Append the token to the response string
                yield response_text                  # Yield the accumulated response so far (for streaming UI update)

    except Exception as e:
        # If any error occurs during the API call, yield an error message
        error_message = f"An error occurred during model inference: {e}"
        print(error_message) # Also print to console for debugging
        yield error_message

# --- Gradio Interface Definition ---

# URL for the header image
header_image_path = "https://cdn-uploads.huggingface.co/production/uploads/6540a02d1389943fef4d2640/j61iZTDaK9g0UW3aWGwWi.gif"

# Ko-fi widget script
kofi_script = """
<script src='https://storage.ko-fi.com/cdn/scripts/overlay-widget.js'></script>
<script>
  kofiWidgetOverlay.draw('sonnydesorbo', {
    'type': 'floating-chat',
    'floating-chat.donateButton.text': 'Support me',
    'floating-chat.donateButton.background-color': '#00b9fe',
    'floating-chat.donateButton.text-color': '#fff'
  });
</script>
"""

# Create a Gradio Blocks layout for more control over the interface
# theme=gr.themes.Soft() applies a soft visual theme
# Add the kofi_script to the head of the HTML page
with gr.Blocks(theme=gr.themes.Soft(), head=kofi_script) as demo:
    # Display an image at the top of the chatbot interface
    gr.Image(
        value=header_image_path, # Source of the image
        label="Chatbot Header",   # Alt text or label (not shown due to show_label=False)
        show_label=False,         # Hide the label text
        interactive=False,        # Make the image non-interactive
        height=100,               # Set the height of the image
        elem_id="chatbot-logo"    # Assign an HTML ID for potential CSS styling
    )

    # Create the chat interface component
    gr.ChatInterface(
        fn=respond,               # The function to call when a message is sent
        chatbot=gr.Chatbot(       # Configure the chatbot display area
            height=500            # Set the height of the chat history display
        ),
        # Additional parameters for ChatInterface can be added here, e.g.:
        # title="Xortron7 Chat",
        # description="Chat with Xortron7, your AI assistant.",
        # examples=[["Hello!", None], ["What is Gradio?", None]],
        # retry_btn=None, # Removes the retry button
        # undo_btn="Delete Previous", # Customizes the undo button
        # clear_btn="Clear Chat", # Customizes the clear button
    )

# --- Application Entry Point ---
if __name__ == "__main__":
    # Launch the Gradio web server
    # show_api=False disables the API documentation page
    # share=False prevents creating a public Gradio link (for local development)
    try:
        demo.launch(show_api=False, share=False)
    except NameError as ne:
        # This might happen if 'client' was not defined due to an error during initialization
        print(f"Gradio demo could not be launched. 'client' might not have been initialized: {ne}")
    except RuntimeError as re:
        # This catches the RuntimeError raised if client initialization failed explicitly
        print(f"Gradio demo could not be launched due to an error during client initialization: {re}")
    except Exception as e:
        print(f"An unexpected error occurred when trying to launch Gradio demo: {e}")