Spaces:
Running
Running
File size: 15,458 Bytes
226ac39 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 |
"""
Code Interpreter Tool
Allows the AI agent to write and execute custom Python code for tasks that don't have predefined tools.
This is what makes it a TRUE AI Agent, not just a function-calling bot.
"""
import os
import sys
import subprocess
import tempfile
from pathlib import Path
from typing import Dict, Any, Optional
import polars as pl
def execute_python_code(
code: str,
working_directory: str = "./outputs/code",
timeout: int = 60,
allow_file_operations: bool = True,
output_file: Optional[str] = None
) -> Dict[str, Any]:
"""
Execute custom Python code written by the AI agent.
This is the KEY tool that transforms the agent from a function-calling bot
into a true AI agent capable of solving ANY data science problem.
Use cases:
- Custom visualizations not covered by existing tools
- Data transformations too specific for generic tools
- Domain-specific calculations
- Interactive dashboards
- Custom export formats
Args:
code: Python code to execute
working_directory: Where to run the code (default: ./outputs/code)
timeout: Maximum execution time in seconds
allow_file_operations: Whether code can read/write files
output_file: Optional file path to save output (e.g., HTML plot)
Returns:
Dict with execution results, stdout, stderr, and any generated files
Example:
# Agent can write custom Plotly code for specific visualizations
code = '''
import plotly.express as px
import pandas as pd
df = pd.read_csv('./temp/sales_data.csv')
fig = px.line(df, x='month', y='sales', color='bike_model',
title='Extended Sales by Month for Each Bike Model')
# Add dropdown filter
fig.update_layout(
updatemenus=[{
'buttons': [{'label': model, 'method': 'update',
'args': [{'visible': [model == m for m in df['bike_model'].unique()]}]}
for model in df['bike_model'].unique()],
'direction': 'down',
'showactive': True
}]
)
fig.write_html('./outputs/code/bike_sales_interactive.html')
print("Chart saved to: ./outputs/code/bike_sales_interactive.html")
'''
result = execute_python_code(code)
"""
try:
# β οΈ CRITICAL: Basic syntax validation BEFORE execution
try:
compile(code, '<string>', 'exec')
except SyntaxError as e:
return {
"success": False,
"error": f"Syntax error in generated code: {str(e)}",
"error_type": "SyntaxError",
"line": e.lineno,
"suggestion": "Fix syntax errors in the code. Common issues: missing quotes, parentheses, indentation"
}
# Create working directory with proper permissions
try:
os.makedirs(working_directory, exist_ok=True)
# Ensure directory is writable
test_file = os.path.join(working_directory, '.write_test')
with open(test_file, 'w') as f:
f.write('test')
os.remove(test_file)
except PermissionError:
return {
"success": False,
"error": f"No write permission for directory: {working_directory}",
"error_type": "PermissionError",
"suggestion": f"Check folder permissions or use a different directory"
}
except Exception as e:
return {
"success": False,
"error": f"Failed to create working directory: {str(e)}",
"error_type": type(e).__name__
}
# Security: Validate code doesn't contain dangerous operations
dangerous_patterns = {
'subprocess': 'Use specialized tools instead of shell commands',
'__import__': 'Dynamic imports not allowed for security',
'eval(': 'eval() is dangerous - rewrite without it',
'exec(': 'exec() is dangerous - rewrite without it',
'compile(': 'compile() not needed - write code directly',
'os.system': 'Shell commands not allowed - use Python libraries',
'os.popen': 'Shell commands not allowed - use Python libraries'
}
for pattern, reason in dangerous_patterns.items():
if pattern in code:
return {
"success": False,
"error": f"Code contains restricted operation: {pattern}",
"error_type": "SecurityError",
"reason": reason,
"suggestion": "Rewrite code using safe Python operations"
}
# Create temporary Python file with better error handling
temp_file = None
try:
with tempfile.NamedTemporaryFile(mode='w', suffix='.py', delete=False,
dir=working_directory, encoding='utf-8') as f:
temp_file = f.name
# Add helper imports at the top + error handling wrapper
enhanced_code = """
# Auto-imported libraries for convenience
import pandas as pd
import polars as pl
import numpy as np
import matplotlib
matplotlib.use('Agg') # Non-interactive backend
import matplotlib.pyplot as plt
import seaborn as sns
import plotly.express as px
import plotly.graph_objects as go
from pathlib import Path
import json
import sys
import traceback
# Ensure output directory exists
import os
os.makedirs('./outputs/code', exist_ok=True)
os.makedirs('./outputs/data', exist_ok=True)
try:
# User's code starts here
""" + "\n".join(" " + line for line in code.split("\n")) + """
except Exception as e:
print(f"β Error in code execution: {str(e)}", file=sys.stderr)
traceback.print_exc()
sys.exit(1)
"""
f.write(enhanced_code)
except Exception as e:
return {
"success": False,
"error": f"Failed to write temporary file: {str(e)}",
"error_type": type(e).__name__,
"suggestion": "Check file write permissions"
}
# Track existing files BEFORE execution to detect new files
existing_files = set()
if allow_file_operations:
for output_dir in ['./outputs/code', './outputs/data', './outputs/plots']:
if os.path.exists(output_dir):
for file_path in Path(output_dir).resolve().glob('**/*'):
if file_path.is_file():
existing_files.add(file_path.resolve())
try:
# Execute the code with better error capture
# Use absolute path and normalize it for Windows
abs_temp_file = os.path.abspath(temp_file)
abs_cwd = os.path.abspath(Path.cwd())
result = subprocess.run(
[sys.executable, abs_temp_file],
capture_output=True,
text=True,
timeout=timeout,
cwd=abs_cwd # Use absolute path to avoid permission issues
)
stdout = result.stdout.strip()
stderr = result.stderr.strip()
returncode = result.returncode
# Check for errors with detailed diagnostics
if returncode != 0:
# Parse error message for common issues
error_hints = []
if "PermissionError" in stderr:
error_hints.append("π‘ File permission issue - check if file is open in another program")
if "FileNotFoundError" in stderr:
error_hints.append("π‘ File not found - check if path is correct (use relative paths like './outputs/data/file.csv')")
if "KeyError" in stderr:
error_hints.append("π‘ Column not found - check column names in the CSV")
if "ModuleNotFoundError" in stderr:
error_hints.append("π‘ Missing library - may need to install additional packages")
if "ValueError" in stderr:
error_hints.append("π‘ Data type mismatch - check data types and conversions")
return {
"success": False,
"error": f"Code execution failed",
"stderr": stderr,
"stdout": stdout if stdout else None,
"error_type": "ExecutionError",
"exit_code": returncode,
"hints": error_hints if error_hints else ["Check the error message above for details"]
}
# Success! Find NEWLY generated files (not existing before execution)
generated_files = []
if allow_file_operations:
cwd = Path.cwd()
for output_dir in ['./outputs/code', './outputs/data', './outputs/plots']:
if os.path.exists(output_dir):
abs_output_dir = Path(output_dir).resolve()
for file_path in abs_output_dir.glob('**/*'):
if file_path.is_file():
abs_file = file_path.resolve()
# Only include if it's NEW (didn't exist before) or MODIFIED
is_new = abs_file not in existing_files
# Check if file was modified in last 5 seconds (just created/updated)
import time
file_age = time.time() - file_path.stat().st_mtime
is_recent = file_age < 5
if (is_new or is_recent):
# Get relative path safely (handle Windows paths)
try:
rel_path = file_path.relative_to(cwd)
except ValueError:
# Fallback: just use the file name with output dir
rel_path = Path(output_dir) / file_path.name
# Only include if not temp file and has content
abs_temp = Path(temp_file).resolve() if temp_file else None
if file_path != abs_temp and file_path.stat().st_size > 0:
generated_files.append(str(rel_path).replace('\\', '/'))
# Sort by modification time (newest first)
if generated_files:
generated_files = sorted(
generated_files,
key=lambda x: Path(x).stat().st_mtime,
reverse=True
)[:10] # Limit to 10 most recent files
return {
"success": True,
"stdout": stdout if stdout else "β
Code executed successfully (no output)",
"stderr": stderr if stderr else None,
"message": "β
Code executed successfully",
"generated_files": generated_files,
"working_directory": working_directory,
"execution_summary": {
"lines_of_code": len(code.split('\n')),
"files_generated": len(generated_files)
}
}
finally:
# Clean up temp file
if temp_file and os.path.exists(temp_file):
try:
os.unlink(temp_file)
except Exception:
pass # Ignore cleanup errors
except subprocess.TimeoutExpired:
return {
"success": False,
"error": f"Code execution timed out after {timeout} seconds",
"error_type": "TimeoutError",
"suggestion": "Code is taking too long. Optimize it or increase timeout. Avoid large loops or heavy computations."
}
except Exception as e:
return {
"success": False,
"error": f"Unexpected error: {str(e)}",
"error_type": type(e).__name__,
"suggestion": "This is an unexpected error. Try simplifying the code."
}
def execute_code_from_file(
file_path: str,
working_directory: str = "./outputs/code",
timeout: int = 60
) -> Dict[str, Any]:
"""
Execute Python code from a file.
Useful when code is too long to pass as a string, or when the agent
wants to run an existing script.
Args:
file_path: Path to Python file to execute
working_directory: Where to run the code
timeout: Maximum execution time in seconds
Returns:
Dict with execution results
"""
try:
# Read code from file
with open(file_path, 'r', encoding='utf-8') as f:
code = f.read()
return execute_python_code(
code=code,
working_directory=working_directory,
timeout=timeout
)
except FileNotFoundError:
return {
"success": False,
"error": f"File not found: {file_path}",
"error_type": "FileNotFoundError"
}
except Exception as e:
return {
"success": False,
"error": f"Failed to read file: {str(e)}",
"error_type": type(e).__name__
}
def generate_custom_visualization(
data_file: str,
visualization_description: str,
output_path: str = "./outputs/code/custom_plot.html",
timeout: int = 60
) -> Dict[str, Any]:
"""
HIGH-LEVEL helper: Generate custom visualization from natural language description.
The agent describes what it wants, and this function attempts to generate the code.
This is a convenience wrapper that could use an LLM to generate the plotting code.
Args:
data_file: Path to dataset
visualization_description: Natural language description of desired plot
output_path: Where to save the visualization
timeout: Execution timeout
Returns:
Dict with execution results
Example:
result = generate_custom_visualization(
data_file="./temp/sales.csv",
visualization_description="Line plot of sales by month for each bike model, with dropdown filter",
output_path="./outputs/code/sales_plot.html"
)
"""
# This is a placeholder - in a full implementation, this would use an LLM
# to generate the Plotly code from the description
return {
"success": False,
"error": "Not yet implemented - use execute_python_code with explicit code instead",
"error_type": "NotImplementedError",
"suggestion": "Write the Plotly code explicitly and use execute_python_code()"
}
|