app.py

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

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", "")

# Initialize search tools with fallback capability
try:
    # Try DuckDuckGo first (default)
    print("Initializing DuckDuckGo search tool...")
    ddg_search_tool = DuckDuckGoSearchTool(max_results=10)
    
    # Test the tool with a simple query
    test_result = ddg_search_tool("test query")
    print("DuckDuckGo search tool initialized successfully.")
    
    # Use DuckDuckGo as the primary search tool
    primary_search_tool = ddg_search_tool
    search_tool_name = "DuckDuckGo"
    
except Exception as e:
    print(f"Error initializing DuckDuckGo search tool: {str(e)}")
    print("Falling back to Google search tool...")
    
    try:
        # Import GoogleSearchTool only if needed
        from smolagents import GoogleSearchTool
        google_search_tool = GoogleSearchTool()
        
        # Test the Google search tool
        test_result = google_search_tool("test query")
        print("Google search tool initialized successfully.")
        
        # Use Google as the fallback search tool
        primary_search_tool = google_search_tool
        search_tool_name = "Google"
        
    except Exception as google_error:
        print(f"Error initializing Google search tool: {str(google_error)}")
        print("WARNING: No working search tool available. Agent functionality will be limited.")
        # Create a minimal replacement that returns an explanatory message
        def search_fallback(query):
            return f"Search functionality unavailable. Both DuckDuckGo and Google search tools failed to initialize. Query was: {query}"
        
        primary_search_tool = search_fallback
        search_tool_name = "Unavailable"

# Initialize the VisitWebpageTool
visit_webpage_tool = VisitWebpageTool()

#@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 parse_json(text: str):
    """
    A safer JSON parser using ast.literal_eval.
    Converts JSON-like strings to Python objects without executing code.
    Handles common JSON literals (true, false, null) by converting them to Python equivalents.
    """
    # Replace JSON literals with Python equivalents
    prepared_text = text.replace("true", "True").replace("false", "False").replace("null", "None")
    
    try:
        import ast
        return ast.literal_eval(prepared_text)
    except (SyntaxError, ValueError) as e:
        raise ValueError(f"Failed to parse JSON: {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. Can be:
            - JSON array of objects (each object becomes a row)
            - Pipe-separated values (first row as headers, subsequent rows as values)
            - Plain text (stored in a single 'text' column)
    
    Returns:
        URL of the created dataset or error message along with the log output.
    """
    log_text = ""
    try:
        # Required imports
        import pandas as pd
        from datasets import Dataset, DatasetDict
        from huggingface_hub import HfApi
        
        # Get API key
        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"
        
        # Set fixed username
        username = "Misfits-and-Machines"
        safe_dataset_name = dataset_name.replace(" ", "_").lower()
        repo_id = f"{username}/{safe_dataset_name}"
        
        log_text += f"Creating dataset: {repo_id}\n"
        
        # Ensure repository exists
        hf_api = HfApi(token=api_key)
        try:
            if not hf_api.repo_exists(repo_id=repo_id, repo_type="dataset"):
                hf_api.create_repo(repo_id=repo_id, repo_type="dataset")
                log_text += f"Created repository: {repo_id}\n"
            else:
                log_text += f"Repository already exists: {repo_id}\n"
        except Exception as e:
            log_text += f"Note when checking/creating repository: {str(e)}\n"
        
        # Process input data
        created_ds = None
        try:
            # Try parsing as JSON using the safer parse_json function
            try:
                json_data = parse_json(conversation_data)
                
                # Process based on data structure
                if isinstance(json_data, list) and all(isinstance(item, dict) for item in json_data):
                    log_text += f"Processing JSON array with {len(json_data)} items\n"
                    
                    # Create a dataset with columns for all keys in the first item
                    # This ensures the dataset structure is consistent
                    first_item = json_data[0]
                    columns = list(first_item.keys())
                    log_text += f"Detected columns: {columns}\n"
                    
                    # Initialize data dictionary with empty lists for each column
                    data_dict = {col: [] for col in columns}
                    
                    # Process each item
                    for item in json_data:
                        for col in columns:
                            # Get the value for this column, or empty string if missing
                            value = item.get(col, "")
                            data_dict[col].append(value)
                    
                    # Debug output to verify data structure
                    for col in columns:
                        log_text += f"Column '{col}' has {len(data_dict[col])} entries\n"
                        
                    # Create dataset from dictionary
                    ds = Dataset.from_dict(data_dict)
                    log_text += f"Created dataset with {len(ds)} rows\n"
                    created_ds = DatasetDict({"train": ds})
                elif isinstance(json_data, dict):
                    log_text += "Processing single JSON object\n"
                    # For a single object, create a dataset with one row
                    data_dict = {k: [v] for k, v in json_data.items()}
                    ds = Dataset.from_dict(data_dict)
                    created_ds = DatasetDict({"train": ds})
                else:
                    raise ValueError("JSON not recognized as array or single object")
            except Exception as json_error:
                log_text += f"Not processing as JSON: {str(json_error)}\n"
                raise json_error  # Propagate to next handler
        except Exception:
            # Try pipe-separated format
            lines = conversation_data.strip().split('\n')
            if '|' in conversation_data and len(lines) > 1:
                log_text += "Processing as pipe-separated data\n"
                headers = [h.strip() for h in lines[0].split('|')]
                log_text += f"Detected headers: {headers}\n"
                
                # Initialize data dictionary
                data_dict = {header: [] for header in headers}
                
                # Process each data row
                for i, line in enumerate(lines[1:], 1):
                    if not line.strip():
                        continue
                    values = [val.strip() for val in line.split('|')]
                    if len(values) == len(headers):
                        for j, header in enumerate(headers):
                            data_dict[header].append(values[j])
                    else:
                        log_text += f"Warning: Skipping row {i} (column count mismatch)\n"
                
                # Create dataset from dictionary
                if all(len(values) > 0 for values in data_dict.values()):
                    ds = Dataset.from_dict(data_dict)
                    log_text += f"Created dataset with {len(ds)} rows\n"
                    created_ds = DatasetDict({"train": ds})
                else:
                    log_text += "No valid rows found in pipe-separated data\n"
                    created_ds = DatasetDict({"train": Dataset.from_dict({"text": [conversation_data]})})
            else:
                # Fallback for plain text
                log_text += "Processing as plain text\n"
                created_ds = DatasetDict({"train": Dataset.from_dict({"text": [conversation_data]})})
        
        # Push using the DatasetDict push_to_hub method.
        log_text += f"Pushing dataset to {repo_id}\n"
        created_ds.push_to_hub(
            repo_id=repo_id,
            token=api_key,
            commit_message=f"Upload dataset: {dataset_name}"
        )
        
        dataset_url = f"https://huggingface.co/datasets/{repo_id}"
        log_text += f"Dataset successfully pushed to: {dataset_url}\n"
        return f"Successfully created dataset at {dataset_url}\nLogs:\n{log_text}"
    except Exception as e:
        import traceback
        error_trace = traceback.format_exc()
        log_text += f"Dataset creation error: {str(e)}\n{error_trace}\n"
        return f"Error creating dataset: {str(e)}\nLogs:\n{log_text}"

@tool
def Dataset_Creator_Tool(dataset_name: str, conversation_data: str) -> str:
    """A tool that creates and pushes a dataset to Hugging Face.
    
    Args:
        dataset_name: Name for the dataset (will be prefixed with 'Misfits-and-Machines/')
        conversation_data: Data content to save in the dataset. Formats supported:
            1. JSON array of objects – Each object becomes a row (keys as columns).
               Example: [{"name": "Product A", "brand": "Company X"}, {"name": "Product B", "brand": "Company Y"}]
            2. Pipe-separated values – First row as headers, remaining rows as values.
               Example: "name | brand\nProduct A | Company X\nProduct B | Company Y"
            3. Plain text – Stored in a single 'text' column.
    
    Returns:
        A link to the created dataset on the Hugging Face Hub or an error message, along with log details.
    """
    try:
        log_text = f"Creating dataset '{dataset_name}' with {len(conversation_data)} characters of data\n"
        log_text += f"Dataset will be created at Misfits-and-Machines/{dataset_name.replace(' ', '_').lower()}\n"
        
        # Call Dataset_Creator_Function directly without trying to define any new functions
        result = Dataset_Creator_Function(dataset_name, conversation_data)
        log_text += f"Dataset creation result: {result}\n"
        
        return log_text
    except Exception as e:
        import traceback
        error_trace = traceback.format_exc()
        return f"Error using Dataset Creator tool: {str(e)}\n{error_trace}"

def verify_dataset_exists(repo_id: str) -> dict:
    """Verify that a dataset exists and is valid on the Hugging Face Hub.
    
    Args:
        repo_id: Full repository ID in format "username/dataset_name"
    
    Returns:
        Dict with "exists" boolean and "message" string
    """
    try:
        # Check if dataset exists using the datasets-server API
        api_url = f"https://datasets-server.huggingface.co/is-valid?dataset={repo_id}"
        response = requests.get(api_url)
        
        # Parse the response
        if response.status_code == 200:
            data = response.json()
            # If any of these are True, the dataset exists in some form
            if data.get("viewer", False) or data.get("preview", False):
                return {"exists": True, "message": "Dataset is valid and accessible"}
            else:
                return {"exists": False, "message": "Dataset exists but may not be fully processed yet"}
        else:
            return {"exists": False, "message": f"API returned status code {response.status_code}"}
    except Exception as e:
        return {"exists": False, "message": f"Error verifying dataset: {str(e)}"}

@tool
def Check_Dataset_Validity(dataset_name: str) -> str:
    """A tool that checks if a dataset exists and is valid on Hugging Face.
    
    Args:
        dataset_name: Name of the dataset to check (with or without organization prefix)
    
    Returns:
        Status message about the dataset validity
    """
    try:
        # Ensure the dataset name has the organization prefix
        if "/" not in dataset_name:
            dataset_name = f"Misfits-and-Machines/{dataset_name.replace(' ', '_').lower()}"
        
        # Check dataset validity
        result = verify_dataset_exists(dataset_name)
        
        if result["exists"]:
            return f"Dataset '{dataset_name}' exists and is valid. You can access it at https://huggingface.co/datasets/{dataset_name}"
        else:
            return f"Dataset '{dataset_name}' could not be verified: {result['message']}. It may still be processing or may not exist."
    except Exception as e:
        return f"Error checking dataset validity: {str(e)}"

@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
)

# Add fallback logic that only activates if the primary model fails
def manage_context(prompt, max_allowed_tokens=30000):
    """Manages large contexts by summarizing or trimming when they get too big.
    
    This helps avoid the 'inputs tokens + max_new_tokens must be <= 32768' error
    by keeping the context size under control.
    
    Args:
        prompt: The full context/prompt that might be too large
        max_allowed_tokens: Maximum number of tokens to allow before trimming
        
    Returns:
        A potentially shortened/summarized version of the prompt
    """
    # Rough token estimation (splitting on spaces is a crude approximation)
    estimated_tokens = len(prompt.split())
    
    # If below threshold, return as is
    if estimated_tokens <= max_allowed_tokens:
        return prompt
    
    print(f"WARNING: Context size ({estimated_tokens} estimated tokens) exceeds limit ({max_allowed_tokens})")
    
    # For extremely large prompts, we need more aggressive handling
    if estimated_tokens > max_allowed_tokens * 1.5:
        print("Performing aggressive context management")
        
        # Approach 1: Keep only the most recent parts of the conversation
        lines = prompt.strip().split('\n')
        
        # Identify structural elements to keep
        instruction_idx = -1
        for i, line in enumerate(lines):
            if "You are a" in line or "I want you to" in line:
                instruction_idx = i
        
        # Always keep the first part with instructions (system prompt)
        keep_beginning = lines[:instruction_idx + 20] if instruction_idx >= 0 else lines[:50]
        
        # Keep the most recent content (approximately half of the max tokens)
        keep_end = lines[-int(max_allowed_tokens/15):]
        
        # Add a note about trimming
        middle_note = [
            "",
            "...",
            "[Context has been trimmed to fit token limits]",
            "...",
            ""
        ]
        
        # Combine parts
        shortened_prompt = "\n".join(keep_beginning + middle_note + keep_end)
        print(f"Context reduced from ~{estimated_tokens} to ~{len(shortened_prompt.split())} estimated tokens")
        return shortened_prompt
    
    # Moderate size reduction for moderately oversized prompts
    else:
        print("Performing moderate context management")
        
        # Split into lines for easier processing
        sections = prompt.split("\n\n")
        
        # Keep important sections like system instructions and recent content
        # Identify which sections to keep or trim
        keep_sections = []
        trim_sections = []
        
        # Process each section
        for i, section in enumerate(sections):
            # Always keep the first few sections (likely instructions)
            if i < 3:
                keep_sections.append(section)
            # Keep the last several sections (most recent and relevant)
            elif i > len(sections) - 8:
                keep_sections.append(section)
            # For code blocks, we should generally keep them
            elif "```" in section:
                keep_sections.append(section)
            # For very short sections, keep them
            elif len(section.split()) < 30:
                keep_sections.append(section)
            # For sections with likely important content, keep them
            elif any(marker in section.lower() for marker in ["important", "key", "critical", "necessary", "must"]):
                keep_sections.append(section)
            # Otherwise, candidate for trimming
            else:
                trim_sections.append(section)
        
        # If we still need to trim more, start removing some of the trim_sections
        if len(" ".join(keep_sections).split()) > max_allowed_tokens * 0.8:
            # Keep only a portion of the trim_sections
            trim_to_keep = int(len(trim_sections) * 0.3)  # Keep 30%
            trim_sections = trim_sections[:trim_to_keep]
        
        # Build final prompt with a note about trimming
        final_sections = keep_sections + ["[Some context has been summarized to fit token limits]"] + trim_sections
        final_prompt = "\n\n".join(final_sections)
        
        print(f"Context reduced from ~{estimated_tokens} to ~{len(final_prompt.split())} estimated tokens")
        return final_prompt


# Now update the try_model_call_with_fallbacks function to use this context management
def try_model_call_with_fallbacks(prompt):
    """Try to use the primary model first with aggressive context management."""
    # First, ALWAYS apply context management, but more aggressively 
    try:
        # Get a rough token count estimate
        estimated_tokens = len(prompt.split())
        print(f"Estimated input tokens: {estimated_tokens}")
        
        # Start with 25000 as the maximum (leaving ~6K tokens buffer for the model limits)
        managed_prompt = manage_context(prompt, max_allowed_tokens=25000)
        
        # If still potentially too large, reduce further
        if len(managed_prompt.split()) > 24000:
            print("First context reduction still too large, reducing further...")
            managed_prompt = manage_context(managed_prompt, max_allowed_tokens=22000)
            
            # Final emergency truncation if needed
            if len(managed_prompt.split()) > 22000:
                print("Emergency truncation required")
                words = managed_prompt.split()
                # Keep first 5000 and last 15000 words with a note in between
                managed_prompt = " ".join(words[:5000]) + "\n\n[CONTEXT SEVERELY TRUNCATED]\n\n" + " ".join(words[-15000:])
        
        print(f"Final managed prompt size: {len(managed_prompt.split())} estimated tokens")
        
        # Temporarily reduce output tokens even further if the prompt is large
        temp_max_tokens = model.max_tokens
        if len(managed_prompt.split()) > 20000:
            print("Large prompt detected, temporarily reducing output tokens")
            model.max_tokens = 750  # Temporarily reduce to 750 for this call
        
        try:
            result = original_call(managed_prompt)
            model.max_tokens = temp_max_tokens  # Restore original setting
            return result
        except Exception as call_error:
            # Restore original setting before handling the error
            model.max_tokens = temp_max_tokens
            raise call_error
            
    except Exception as primary_error:
        # If we still get a token limit error, try even more aggressive reduction
        if "Input validation error: inputs tokens + max_new_tokens" in str(primary_error):
            try:
                print("Critical: Token limit exceeded despite context management. Implementing emergency measures...")
                # Take a more drastic approach - keep only system instructions and last part
                lines = prompt.strip().split('\n')
                # Keep first 50 lines and last 100 lines only
                emergency_prompt = "\n".join(lines[:50] + ["\n[MAJORITY OF CONTEXT REMOVED DUE TO TOKEN LIMITS]\n"] + lines[-100:])
                
                # Reduce output tokens drastically
                temp_max_tokens = model.max_tokens
                model.max_tokens = 500
                try:
                    result = original_call(emergency_prompt)
                    model.max_tokens = temp_max_tokens
                    return result
                except Exception:
                    model.max_tokens = temp_max_tokens
                    print("Emergency measures failed. Trying fallback models...")
            except Exception:
                print("Emergency context management failed. Proceeding to fallback models...")
        
        print(f"Primary model call failed: {str(primary_error)}")
        print("Trying fallback models...")
        
        # Rest of fallback logic remains the same...
        # List of fallback models
        fallbacks = [
            {
                "provider": "sambanova",
                "model_name": "Qwen/Qwen2.5-Coder-32B-Instruct",
                "display_name": "Qwen 2.5 Coder 32B"
            },
            {
                "provider": "hf-inference",
                "model_name": "deepseek-ai/DeepSeek-R1-Distill-Qwen-32B",
                "display_name": "DeepSeek R1 Distill Qwen 32B"
            }
        ]
        
        # Get API key
        api_key = os.getenv("HF_API_KEY") or os.getenv("HUGGINGFACE_API_KEY")
        if not api_key:
            raise ValueError("No Hugging Face API key found in environment variables")
        
        # Try each fallback model in sequence with highly aggressive context management
        for fallback in fallbacks:
            try:
                print(f"Trying fallback model: {fallback['display_name']}")
                client = InferenceClient(provider=fallback["provider"], api_key=api_key)
                
                # Apply even more aggressive context management for fallbacks
                emergency_prompt = manage_context(prompt, max_allowed_tokens=15000)
                messages = [{"role": "user", "content": emergency_prompt}]
                
                completion = client.chat.completions.create(
                    model=fallback["model_name"],
                    messages=messages,
                    max_tokens=1000,  # Reduced tokens for output
                    temperature=0.5
                )
                print(f"Successfully used fallback model: {fallback['display_name']}")
                return completion.choices[0].message.content
            except Exception as e:
                print(f"Fallback model {fallback['display_name']} failed: {str(e)}")
                continue
        
        # If all fallbacks fail, provide a useful error message
        return "ERROR: Unable to process request due to context size limitations. Please break your request into smaller parts or simplify your query."

# Monkey patch the model's __call__ method to use our fallback logic
original_call = model.__call__
model.__call__ = try_model_call_with_fallbacks

# Reduce the model's output tokens immediately to improve chances of success
model.max_tokens = 1000  # Reduce from 2096 to 1000 to stay under token limits

# 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)

# Override CodeAgent run to automatically clear context periodically
original_code_agent_run = CodeAgent.run

def context_managed_run(self, query, max_steps=None, persist_conversation=None):
    """Override to periodically clear conversation context after steps"""
    # Initialize step counter if not present
    if not hasattr(self, '_context_step_counter'):
        self._context_step_counter = 0
    
    # Increment counter
    self._context_step_counter += 1
    
    # Every 5 steps, perform aggressive context management
    if self._context_step_counter >= 5:
        print("Performing periodic conversation cleanup")
        self._context_step_counter = 0
        
        # Clear most of the conversation history while keeping essential parts
        if hasattr(self, 'conversation') and len(self.conversation) > 8:
            original_length = len(self.conversation)
            
            # Find system messages
            system_messages = [msg for msg in self.conversation if msg.get('role') == 'system']
            
            # Get the most recent exchanges (last 4 pairs of messages)
            recent_messages = self.conversation[-8:]
            
            # Rebuild conversation with system messages + note + recent messages
            self.conversation = system_messages + [
                {'role': 'system', 'content': '[NOTICE: Previous conversation history has been trimmed to manage context size]'}
            ] + recent_messages
            
            print(f"Conversation history trimmed from {original_length} to {len(self.conversation)} messages")
    
    # Call the original run method
    return original_code_agent_run(self, query, max_steps, persist_conversation)

# Apply the monkey patch to CodeAgent class
CodeAgent.run = context_managed_run
    
# Initialize the agent using standard smolagents patterns
agent = CodeAgent(
    model=model,
    tools=[
        final_answer,
        Sonar_Web_Search_Tool,
        primary_search_tool,  # This is already set to either DuckDuckGo, Google, or fallback
        get_current_time_in_timezone,
        image_generation_tool,
        Dataset_Creator_Tool,
        Check_Dataset_Validity,
        visit_webpage_tool,  # This is correctly initialized as VisitWebpageTool()
    ],
    max_steps=12,
    verbosity_level=1,
    grammar=None,
    planning_interval=2,
    name="Research Assistant",
    description="""An AI assistant that can search the web, create datasets, and answer questions # Note about working within token limits
                # When using with queries that might exceed token limits, consider:
                # 1. Breaking tasks into smaller sub-tasks
                # 2. Limiting the amount of data returned by search tools
                # 3. Using the planning_interval to enable more effective reasoning""",
    prompt_templates=prompt_templates
)

# Add informative message about which search tool is being used
print(f"Agent initialized with {search_tool_name} as primary search tool")
print(f"Available tools: final_answer, Sonar_Web_Search_Tool, {search_tool_name}, get_current_time_in_timezone, image_generation_tool, Dataset_Creator_Tool, Check_Dataset_Validity, visit_webpage_tool")

# Note about working within token limits - add this comment
# When using with queries that might exceed token limits, consider:
# 1. Breaking tasks into smaller sub-tasks
# 2. Limiting the amount of data returned by search tools
# 3. Using the planning_interval to enable more effective reasoning

# 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