app.py

from smolagents import (
    CodeAgent,
    HfApiModel,
    TransformersModel,
    OpenAIServerModel,
    load_tool,
    tool,
)
import os
import datetime
import pytz
import yaml
from tools import (
    FinalAnswerTool,
    VisitWebpageTool,
    DuckDuckGoSearchTool,  # This is also built-in in smolagents
)


# 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)}"


final_answer = FinalAnswerTool()

if IS_IN_HF_SPACE := os.getenv("SPACE_ID"):
    # model_id = (
    #     # BUG: Seems we are failed to call this model => bad request
    #     "https://wxknx1kg971u7k1n.us-east-1.aws.endpoints.huggingface.cloud"  # it is possible that this model may be overloaded
    #     if True
    #     # BUG: Model not loaded on the server: https://api-inference.huggingface.co/models/Qwen/Qwen2.5-Coder-32B-Instruct/v1/chat/completions. Please retry with a higher timeout (current: 120)
    #     # BUG: TooManyRequests: Please log in or use a HF access token
    #     else "Qwen/Qwen2.5-Coder-32B-Instruct"  # The default value of HfApiModel
    # )

    # https://discord.com/channels/879548962464493619/1336751588391391283/1338901490651103414
    model_id = "deepseek-ai/DeepSeek-R1-Distill-Qwen-32B"

    print("Using HfApiModel with model_id:", model_id)

    model = HfApiModel(
        max_tokens=2096 if IS_IN_HF_SPACE else None,
        temperature=0.5,
        model_id=model_id,
        custom_role_conversions=None,
    )

else:
    from dotenv import load_dotenv

    # NOTE: to load the Hugging Face API Key
    curr_dir = os.path.dirname(os.path.abspath(__file__))
    load_dotenv(os.path.join(curr_dir, "../../.env"))

    if OPENAI_API_KEY := os.getenv("OPENAI_API_KEY"):
        print("Using OpenAIServerModel with model_id: gpt-3.5-turbo")

        model = OpenAIServerModel(model_id="gpt-3.5-turbo", api_key=OPENAI_API_KEY)
    else:
        # NOTE: this model is not good enough for agent (and still might be too heavy to run on a normal computer)
        # print(
        #     "Using TransformersModel with model_id: HuggingFaceTB/SmolLM2-1.7B-Instruct"
        # )
        #
        # model = TransformersModel(
        #     model_id="HuggingFaceTB/SmolLM2-1.7B-Instruct", trust_remote_code=True
        # )

        # NOTE: this model is well enough to use simple tools (but need authentication)
        # https://huggingface.co/meta-llama/Llama-3.2-11B-Vision-Instruct
        # https://huggingface.co/settings/gated-repos
        print(
            "Using HfApiModel with model_id: meta-llama/Llama-3.2-11B-Vision-Instruct"
        )

        # NOTE: this use InferenceClient under the hood
        model = HfApiModel(
            temperature=0.5,
            model_id="meta-llama/Llama-3.2-11B-Vision-Instruct",
            custom_role_conversions=None,
        )

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

duckduckgo_tool = DuckDuckGoSearchTool()
visit_webpage_tool = VisitWebpageTool()

with open("prompts.yaml", "r") as stream:
    prompt_templates = yaml.safe_load(stream)

agent = CodeAgent(
    model=model,
    tools=[
        final_answer,
        my_custom_tool,
        get_current_time_in_timezone,
        image_generation_tool,
        duckduckgo_tool,
        visit_webpage_tool,
    ],  # Add your tools here (don't remove final answer)
    max_steps=6,
    verbosity_level=1,
    grammar=None,
    planning_interval=None,
    name=None,
    description=None,
    prompt_templates=prompt_templates,
)

if __name__ == "__main__":
    from Gradio_UI import GradioUI

    GradioUI(agent).launch()