∞ Regex Creator

Documentation

• Class name: LLMRegexCreator
• Category: SALT/Language Toolkit/Tools/Regex
• Output node: False

The LLMRegexCreator node is designed to generate regular expression (regex) patterns based on a given description and optional additional directions. It utilizes a language model to interpret the requirements and produce a well-formed regex pattern that accurately matches the specified criteria.

Input types

Required

• llm_model
  • Specifies the language model to use for generating the regex pattern. It is crucial for interpreting the description and extra directions to create the regex.
  • Comfy dtype: LLM_MODEL
  • Python dtype: Dict[str, Any]
• description
  • The primary description of the regex pattern to be created. This description is the basis for the regex generation, guiding the language model in crafting the pattern.
  • Comfy dtype: STRING
  • Python dtype: str

Optional

• extra_directions
  • Optional additional instructions or constraints to guide the language model in generating the regex pattern. These directions can refine the output by providing more context or specific requirements.
  • Comfy dtype: STRING
  • Python dtype: str

Output types

• regex_pattern
  • Comfy dtype: STRING
  • The generated regex pattern based on the provided description and optional extra directions.
  • Python dtype: str

Usage tips

• Infra type: CPU
• Common nodes: unknown

Source code

class LLMRegexCreator:
    @classmethod
    def INPUT_TYPES(cls):
        return {
            "required": {
                "llm_model": ("LLM_MODEL",),
                "description": ("STRING", {"multiline": True, "dynamicPrompts": False, "placeholder": "Describe regex pattern to create, optionally provide example"}),
            },
            "optional": {
                "extra_directions": ("STRING", {"multiline": True, "dynamicPrompts": False, "placeholder": "Extra directions for the LLM to follow..."}),
            }
        }

    RETURN_TYPES = ("STRING",)
    RETURN_NAMES = ("regex_pattern",)

    FUNCTION = "create_regex"
    CATEGORY = f"{MENU_NAME}/{SUB_MENU_NAME}/Tools/Regex"

    def create_regex(self, llm_model, description, extra_directions=""):
        prompt = (
            f"Create only a well formed regex pattern based on the following description:\n\n"
            f"{description}\n\n"
            f"{extra_directions}\n\n"
            "Please ensure the regex pattern is concise and accurately matches the described criteria."
        )

        response = llm_model['llm'].complete(prompt)

        return (response.text,)