Build AI Apps with Python: Full-Featured CLI Agent — Tools Guardrails History | Episode 23

Build AI Apps with Python: Full-Featured CLI Agent — Tools, Guardrails, History

This example demonstrates how to build a command-line AI assistant in Python that integrates multiple tools, enforces input guardrails, and maintains conversation history. The agent supports math calculations, word definitions, and weather queries, while preventing unsafe requests from reaching the AI model.

Code

import sys

# Tool functions
def calculate(expression):
  # Simple eval for math expressions (use cautiously in real apps)
  try:
    return str(eval(expression, {"__builtins__": {}}))
  except Exception as e:
    return f"Error calculating expression: {e}"

def define_word(word):
  # Dummy dictionary lookup
  definitions = {
    "python": "A high-level programming language.",
    "ai": "Artificial intelligence, the simulation of human intelligence by machines."
  }
  return definitions.get(word.lower(), "Definition not found.")

def get_weather(location):
  # Placeholder weather info
  return f"The weather in {location} is sunny and 75°F."

# Guardrail to block dangerous inputs
def guardrails(user_input):
  blocked_keywords = ["rm -rf", "shutdown", "delete", "format", "drop table"]
  for keyword in blocked_keywords:
    if keyword in user_input.lower():
      return False, f"Request blocked due to unsafe keyword: '{keyword}'"
  return True, ""

# Conversation history storage
history = []

def main():
  print("Welcome to the AI CLI Agent. Type 'quit' to exit.")
  while True:
    user_input = input("You: ").strip()
    if user_input.lower() == "quit":
      print("Goodbye!")
      break

    allowed, message = guardrails(user_input)
    if not allowed:
      print(f"Guardrails: {message}")
      continue

    # Check for tool usage commands
    if user_input.startswith("calculate "):
      expr = user_input[len("calculate "):]
      response = calculate(expr)
    elif user_input.startswith("define_word "):
      word = user_input[len("define_word "):]
      response = define_word(word)
    elif user_input.startswith("get_weather "):
      location = user_input[len("get_weather "):]
      response = get_weather(location)
    else:
      # Default AI response placeholder
      response = f"AI response to: {user_input}"

    # Save to history
    history.append({"user": user_input, "agent": response})

    print(f"Agent: {response}")

if __name__ == "__main__":
  main()

Key Points

• The agent uses separate tool functions to handle math, definitions, and weather queries.
• Input guardrails scan user input for dangerous commands and block them before processing.
• Conversation history is stored in a list to maintain context across turns.
• An interactive CLI loop lets users chat, call tools, and quit gracefully.
• This pattern combines multiple AI agent features into a cohesive terminal assistant.