app.py

from smolagents import CodeAgent,DuckDuckGoSearchTool, HfApiModel,load_tool,tool
import datetime
import requests
import pytz
import yaml
from tools.final_answer import FinalAnswerTool

from Gradio_UI import GradioUI

# Below is an example of a tool that does nothing. Amaze us with your creativity !
@tool
def my_custom_tool(arg1:str, arg2:int)-> str: #it's import to specify the return type
    #Keep this format for the description / args / args description but feel free to modify the tool
    """A tool that does nothing yet 
    Args:
        arg1: the first argument
        arg2: the second argument
    """
    return "What magic will you build ?"

@tool
def get_current_time_in_timezone(timezone: str) -> str:
    """A tool that fetches the current local time in a specified timezone.
    Args:
        timezone: A string representing a valid timezone (e.g., 'America/New_York').
    """
    try:
        # Create timezone object
        tz = pytz.timezone(timezone)
        # Get current time in that timezone
        local_time = datetime.datetime.now(tz).strftime("%Y-%m-%d %H:%M:%S")
        return f"The current local time in {timezone} is: {local_time}"
    except Exception as e:
        return f"Error fetching time for timezone '{timezone}': {str(e)}"

@tool
def create_prompt_for_image_generation(user_prompt: str) -> str:

    """Executes a prompt using a language model to create a detaled prompt
    for image generation based on user_prompt. Returns prompt for image_generation_tool.

    Args:
        user_prompt: A string - the user's text prompt (e.g. 'Giraffe in Louvre in front of Mona Lisa Painting by Leonardo'.
    Output type: str
    """
    # Prompt parts
    prefix="Generate a detailed and structured FLUX-Schnell-compatible prompt based on the following short description of an image: "
    postfix="""
    The generated prompt should follow these guidelines:
	1.	Foreground, Middle Ground, and Background: Clearly describe elements in each layer of the image in an organized manner.
	2.	Tone and Style: Specify the tone (e.g., cinematic, surreal, vibrant) and artistic style (e.g., photorealistic, painterly, abstract).
	3.	Color Palette: Include details about the dominant colors or overall color scheme.
	4.	Perspective and Camera Details: Mention the point of view (e.g., wide-angle, close-up), camera type, lens, aperture, and lighting conditions if applicable.
	5.	Additional Details: Highlight any specific objects, text, or unique features with clear emphasis (e.g., 'with green text' or 'emphasis on golden hour lighting').
	6.	Output Settings: Suggest aspect ratio, output format (e.g., JPG), quality level, and seed for reproducibility.
    Ensure that the generated prompt is logical, descriptive, and written in natural language to maximize compatibility with FLUX-Schnell capabilities.
    Example Input:
    An image of a serene forest with a small cabin.
    Example Output:
    In the foreground, a lush green forest floor covered with moss and scattered wildflowers. 
    In the middle ground, a cozy wooden cabin with smoke gently rising from its chimney. 
    In the background, towering pine trees fading into a misty horizon. 
    The tone is tranquil and inviting, with a photorealistic style. 
    The color palette includes rich greens, warm browns for the cabin, and soft gray mist. 
    The perspective is slightly elevated as if viewed from a drone camera at sunrise, 
    capturing golden hour lighting for soft shadows and warm highlights. 
    The aspect ratio is 1:1, output format JPG, high quality, using seed 42 for reproducibility.

    Output only the final prompt without any comments or introduction.
    """
    model = HfApiModel(
    max_tokens=384,
    temperature=1.0,
    model_id='Qwen/Qwen2.5-Coder-32B-Instruct',
    # custom_role_conversions=None,
    )
    
    prompt = prefix + user_prompt + '. ' + postfix 
    messages = [{"role": "user", "content": prompt}]
    
    try:
        # response = model(
        #     prompt=prompt, temperature=1., max_tokens=512)
        response = model(messages, stop_sequences=["END"])
        # return response['choices'][0]['text'] 
        # return response['choices'][0]['message']['content']
        print(response.content)
        return response.content

    except Exception as e:
        print(f"Error during LLM call: {str(e)}")
        return f"Error during LLM call: {str(e)}"


final_answer = FinalAnswerTool()

# If the agent does not answer, the model is overloaded, please use another model or the following Hugging Face Endpoint that also contains qwen2.5 coder:
# model_id='https://pflgm2locj2t89co.us-east-1.aws.endpoints.huggingface.cloud' 

model = HfApiModel(
max_tokens=2096,
temperature=0.5,
model_id='Qwen/Qwen2.5-Coder-32B-Instruct',# it is possible that this model may be overloaded
custom_role_conversions=None,
)


# Import tool from Hub
image_generation_tool = load_tool("agents-course/text-to-image", trust_remote_code=True)

with open("prompts.yaml", 'r') as stream:
    prompt_templates = yaml.safe_load(stream)
    
agent = CodeAgent(
    model=model,
    tools=[final_answer, create_prompt_for_image_generation, image_generation_tool], ## add your tools here (don't remove final answer)
    max_steps=6,
    verbosity_level=1,
    grammar=None,
    planning_interval=None,
    name="Agent-Unit1",
    description=None,
    prompt_templates=prompt_templates
)


GradioUI(agent).launch()