utils.py

from llm_guard.input_scanners import PromptInjection
from llm_guard.input_scanners.prompt_injection import MatchType


def is_subsequence(main_string, sub_string):
    """
    Checks if sub_string is a subsequence of main_string.
    A subsequence allows arbitrary characters in between the characters of sub_string in main_string.

    Parameters:
    main_string (str): The string in which to search.
    sub_string (str): The string to search for.

    Returns:
    bool: True if sub_string is a subsequence of main_string, False otherwise.
    """
    main_string = main_string.lower()
    sub_string = sub_string.lower()

    main_len = len(main_string)
    sub_len = len(sub_string)

    if sub_len == 0:
        return True
    if main_len == 0:
        return False

    main_index = 0
    sub_index = 0

    while main_index < main_len and sub_index < sub_len:
        if main_string[main_index] == sub_string[sub_index]:
            sub_index += 1
        main_index += 1

    return sub_index == sub_len


def is_malicious(user_input: str):
    """Function to check if the user input is malicious. Uses the LLM Guard prompt injection scanner.
    Uses a special model `ProtectAI/deberta-v3-base-prompt-injection-v2` to scan the user input for malicious content.
    """
    scanner = PromptInjection(threshold=0.5, match_type=MatchType.FULL)
    sanitized_prompt, is_valid, risk_score = scanner.scan(user_input)
    return is_valid, risk_score