app.py

from smolagents import CodeAgent, DuckDuckGoSearchTool, HfApiModel, load_tool, tool
import datetime
import requests
import pytz
import yaml
import os
import json
import uuid
from datasets import Dataset
from huggingface_hub import HfApi
from openai import OpenAI
from tools.final_answer import FinalAnswerTool

from Gradio_UI import GradioUI

# Define the Perplexity system prompt
Perplex_Assistant_Prompt = """You are a helpful AI assistant that searches the web for accurate information."""

# Set up API key in environment variable as expected by HfApiModel
os.environ["HUGGINGFACE_API_TOKEN"] = os.getenv("HUGGINGFACE_API_KEY", "")

#@weave.op()
def tracked_perplexity_call(prompt: str, system_messages: str, model_name: str = "sonar-pro", assistant_meta: bool = False):
    """Enhanced Perplexity API call with explicit model tracking."""
    client = OpenAI(api_key=os.getenv("PERPLEXITY_API_KEY"), base_url="https://api.perplexity.ai")
    system_message = Perplex_Assistant_Prompt
    if assistant_meta:
        system_message += f"\n\n{system_messages}"
    
    # Minimal parameters for Perplexity
    return client.chat.completions.create(
        model=model_name,
        messages=[
            {"role": "system", "content": system_message},
            {"role": "user", "content": prompt},
        ],
        stream=False,
    ).choices[0].message.content

@tool
def Sonar_Web_Search_Tool(arg1: str, arg2: str) -> str:
    """A tool that accesses Perplexity Sonar to search the web when the answer requires or would benefit from a real world web reference. 
    Args:
        arg1: User Prompt
        arg2: Details on the desired web search results as system message for sonar web search
    """
    try:
        sonar_response = tracked_perplexity_call(arg1, arg2)
        return sonar_response
    except Exception as e:
        return f"Error using Sonar Websearch tool '{arg1} {arg2}': {str(e)}"
    

def Dataset_Creator_Function(dataset_name: str, conversation_data: str) -> str:
    """Creates and pushes a dataset to Hugging Face with the conversation history.
    
    Args:
        dataset_name: Name for the dataset (will be prefixed with username)
        conversation_data: String representing the conversation data
    
    Returns:
        URL of the created dataset or error message
    """
    try:
        # Get API key from environment variables
        api_key = os.getenv("HF_API_KEY") or os.getenv("HUGGINGFACE_API_KEY", "")
        if not api_key:
            return "Error: No Hugging Face API key found in environment variables. Please set HF_API_KEY or HUGGINGFACE_API_KEY."
            
        # Force the username to be the known value
        username = "Misfits-and-Machines"
        
        # Initialize Hugging Face API
        hf_api = HfApi(token=api_key)
        
        # Sanitize dataset name - use underscores instead of dashes for better compatibility
        safe_dataset_name = dataset_name.replace(" ", "_").lower()
        repo_id = f"{username}/{safe_dataset_name}"
        
        print(f"Creating dataset repository: {repo_id}")
        
        # Prepare dataset with appropriate structure
        # First, ensure we have a proper train split with necessary fields
        dataset_dict = {
            "text": [conversation_data],
            "timestamp": [datetime.datetime.now().isoformat()],
            "dataset_id": [str(uuid.uuid4())]
        }
        
        # Create a Hugging Face dataset
        dataset = Dataset.from_dict(dataset_dict)
        
        # Standard practice is to have a train split for datasets
        dataset_dict = {"train": dataset}
        
        # Check if the repository already exists
        try:
            repo_exists = hf_api.repo_exists(repo_id=repo_id, repo_type="dataset")
            if repo_exists:
                print(f"Repository {repo_id} already exists")
            else:
                # Create repo if it doesn't exist
                hf_api.create_repo(repo_id=repo_id, repo_type="dataset", exist_ok=True)
                print(f"Repository {repo_id} created successfully")
        except Exception as repo_error:
            print(f"Repository check/creation error: {str(repo_error)}")
            # Continue anyway as push_to_hub might create the repo
        
        # Push dataset to the Hub with appropriate parameters
        print(f"Pushing dataset to {repo_id}")
        
        # Create URL for monitoring - we'll show this to the user so they can check progress
        dataset_url = f"https://huggingface.co/datasets/{repo_id}"
        print(f"Dataset URL will be: {dataset_url}")
        
        # Push with careful parameter selection
        dataset.push_to_hub(
            repo_id=repo_id,
            token=api_key,
            split="train",  # Use a proper split name
            commit_message=f"Upload dataset: {dataset_name}"
        )
        
        print(f"Dataset successfully pushed to: {dataset_url}")
        return f"Successfully created dataset at {dataset_url} - please check this URL to verify your dataset is visible"
    except Exception as e:
        import traceback
        error_trace = traceback.format_exc()
        print(f"Dataset creation error: {str(e)}\n{error_trace}")
        return f"Error creating dataset: {str(e)}\n\nTo troubleshoot:\n1. Verify API key is valid\n2. Try with a different dataset name\n3. Check if you have write permissions for the Misfits-and-Machines organization"

@tool
def Dataset_Creator_Tool(dataset_name: str, conversation_data: str) -> str:
    """A tool that posts a new dataset of the current conversation to Hugging Face.
    
    Args:
        dataset_name: Name for the dataset (will be prefixed with 'Misfits-and-Machines/')
        conversation_data: String content to save to the dataset (no JSON conversion needed)
    
    Returns:
        Link to the created dataset or error message with troubleshooting steps
    """
    try:
        print(f"Creating dataset '{dataset_name}' with {len(conversation_data)} characters of data")
        print(f"Dataset will be created at Misfits-and-Machines/{dataset_name.replace(' ', '_').lower()}")
        result = Dataset_Creator_Function(dataset_name, conversation_data)
        print(f"Dataset creation result: {result}")
        return result
    except Exception as e:
        import traceback
        error_trace = traceback.format_exc()
        return f"Error using Dataset Creator tool: {str(e)}\n{error_trace}\n\nPlease try with a simpler dataset name using only letters, numbers and underscores."


@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()

# Remove the huggingface_api_key parameter - it's not supported
model = HfApiModel(
    max_tokens=2096,
    temperature=0.5,
    model_id='https://pflgm2locj2t89co.us-east-1.aws.endpoints.huggingface.cloud',  # Using the backup endpoint
    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, 
        Sonar_Web_Search_Tool, 
        get_current_time_in_timezone, 
        image_generation_tool,
        Dataset_Creator_Tool
    ],  
    max_steps=6,
    verbosity_level=1,
    grammar=None,
    planning_interval=None,
    name=None,
    description=None,
    prompt_templates=prompt_templates
)

# To fix the TypeError in Gradio_UI.py, you would need to modify that file
# For now, we'll just use the agent directly
try:
    GradioUI(agent).launch()
except TypeError as e:
    if "unsupported operand type(s) for +=" in str(e):
        print("Error: Token counting issue in Gradio UI")
        print("To fix, edit Gradio_UI.py and change:")
        print("total_input_tokens += agent.model.last_input_token_count")
        print("To:")
        print("total_input_tokens += (agent.model.last_input_token_count or 0)")
    else:
        raise e