∞ JSON Composer

Documentation

• Class name: LLMJsonComposer
• Category: SALT/Language Toolkit/Tools/JSON
• Output node: False

The LLMJsonComposer node is designed to generate valid JSON objects from given text inputs by leveraging a language model. It processes textual data along with a set of classifiers and optional directions to compose JSON structures that encapsulate the provided information accurately and coherently.

Input types

Required

• llm_model
  • Specifies the language model to be used for generating the JSON object, playing a crucial role in interpreting the input text and classifiers to produce the desired output.
  • Comfy dtype: LLM_MODEL
  • Python dtype: dict
• text_input
  • The primary textual data from which the JSON object will be composed, serving as the base content for the generation process.
  • Comfy dtype: STRING
  • Python dtype: str
• classifier_list
  • A comma-separated list of classifiers that guide the JSON composition by categorizing the input text, influencing the structure and content of the generated JSON.
  • Comfy dtype: STRING
  • Python dtype: str

Optional

• extra_directions
  • Optional additional instructions for the language model to follow when composing the JSON, allowing for more tailored and specific outputs.
  • Comfy dtype: STRING
  • Python dtype: str

Output types

• json_output
  • Comfy dtype: STRING
  • The generated JSON object as a string, representing a structured and valid JSON format of the input text and classifiers.
  • Python dtype: str

Usage tips

• Infra type: GPU
• Common nodes: unknown

Source code

class LLMJsonComposer:
    @classmethod
    def INPUT_TYPES(cls):
        return {
            "required": {
                "llm_model": ("LLM_MODEL",),
                "text_input": ("STRING", {"multiline": True, "dynamicPrompts": False, "placeholder": "Data..."}),
                "classifier_list": ("STRING", {"multiline": False, "dynamicPrompts": False}),
            },
            "optional": {
                "extra_directions": ("STRING", {"multiline": True, "dynamicPrompts": False, "placeholder": "Extra directions for the LLM to follow..."}),
            }
        }

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

    FUNCTION = "compose_json"
    CATEGORY = f"{MENU_NAME}/{SUB_MENU_NAME}/Tools/JSON"

    def compose_json(self, llm_model, text_input, classifier_list, extra_directions=""):
        classifier_list = [item.strip() for item in classifier_list.split(",") if item.strip()]
        prompt = f"{text_input}\n\n###\n\nGiven the above text, create a valid JSON object utilizing *all* of the data; using the following classifiers: {classifier_list}.\n\n{extra_directions}\n\nPlease ensure the JSON output is properly formatted, and does not omit any data."
        response = llm_model['llm'].complete(prompt)
        return (response.text,)