Categorizer.py was deleted, but __init__.py still listed it, so every ComfyUI startup printed an import traceback and "Could not load Categorizer". Also drops it from docs/ai_agents.md and syncs __version__. Version 2.10.3. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_013Lm4YDXhwj49hAA1RKS58Q
7.8 KiB
🤖 Master Key
A powerful "Idea Engine" that acts as the creative starting point for generating new and complex prompts.
Philosophy & Use Case The Master Key is your creative spark plug. It's designed to be the very first step in your prompt generation workflow, providing a clean and focused space to capture your initial idea before it's sent to a powerful AI for expansion. It acts as the "mission briefing" for your AI Art Director.
It's perfect for two main scenarios:
Keyword Expansion: You provide a handful of simple keywords (e.g., "boy, dog, park, red ball"), and the Master Key passes them along to an AI that can weave them into a coherent and detailed scene.
Concept Expansion: You provide a simple sentence describing a scene (e.g., "create a prompt about a boy and a dog playing with a red ball in the park"), and the Master Key feeds this core concept to an AI to be fleshed out with cinematic details, emotional depth, and artistic style.
Inputs This node has no wired inputs. It is a "start node" where you type your initial idea directly into its text box.
Outputs text (STRING): The raw text from the input widget, passed through to be used as the prompt for another node (typically the Gemini API node).
Example Workflow The Master Key is designed to be connected directly to the text or prompt input of an AI node, like the Gemini API node. This allows the AI to take the simple idea from the Master Key and transform it into a rich, detailed final prompt.
🤖 Gemini API
The core engine that connects your workflow to Google's powerful Gemini family of models, allowing for advanced text and image analysis and generation.
Philosophy & Use Case This node is the foundational "brain" for all AI-powered tasks in the Creepybits suite. It takes a complete creative brief—typically from a node like Master Key or Director's Slate—and sends it along with a set of fine-tuning parameters to the Gemini model for processing. The model then "thinks" about the request and sends back a newly generated, enhanced, or analyzed text.
You will almost always use this node as the second step in a prompt-generation workflow, right after a node that prepares your initial idea. It is the essential component that brings the AI's intelligence into your creative process.
Inputs & Parameters image (IMAGE, optional): An input image for the AI to analyze, allowing for powerful multimodal workflows (e.g., describing a picture or using it as inspiration).
system_prompt (STRING): This is the primary mission briefing for the AI. It should contain the user's core idea and any guiding instructions, typically piped in from the text output of a Master Key node.
Model Parameters (model, temperature, top_p, etc.): These sliders and dropdowns are the fine-tuning controls for the AI's "personality." They allow you to control everything from its creativity (temperature) to the length of its response (max_output_tokens).
API Key
- Navigate to custom_nodes/Creepy_nodes/assets/scripts/.
- Paste your key into gemini_api_key_free.txt (and/or gemini_api_key_paid.txt).
- That's it. api_keys_config.json already points at those two files. To add more keys, add entries to it; paths can be absolute or relative to the scripts folder:
{
"Gemini Free": "gemini_api_key_free.txt",
"Gemini Paid": "C:\Path\To\gemini_api_key_paid.txt"
}
Alternatively, set the GEMINI_API_KEY environment variable and pick "Environment (GEMINI_API_KEY)" in the api_key_selection dropdown. To keep api_keys_config.json somewhere else, set CREEPYBITS_API_CONFIG to its full path.
The model dropdown lists current Gemini models. To use a model that isn't listed (for example a newly released one), type its ID into custom_model.
Note: The user_instructions text box is a legacy field and is not used in current workflows. It can be safely ignored.
Outputs text (STRING): The final text generated by the Gemini model, ready to be used as a prompt for a text-to-image model or for any other purpose in your workflow.
Example Workflow The Gemini API node acts as the bridge between your initial idea and the final, AI-generated prompt. Here is its most common use case:
The user types a simple idea into the Master Key node.
The Master Key sends that idea to the system_prompt input of the Gemini API node.
The Gemini API node sends the idea to the external AI for processing.
The AI generates a rich, detailed prompt and sends it back to the Gemini API node's text output.
🤖 The AI Creative Council (Prompt Helpers)
A suite of specialized AI "persona" nodes designed to shape, analyze, and refine your ideas before they are sent to the main generation engine.
Philosophy & Use Case The Creative Council is a team of specialized AI experts. While nodes like Master Key are for generating entirely new concepts from scratch, these nodes are for refining an existing idea. Each node acts like a different specialist you bring in to consult on your prompt. Do you need an art critic? A summarizer? A keyword expert? This suite provides the right brain for the right job, acting as a powerful pre-processing step to create more intelligent and targeted final prompts.
The Suite Includes This collection of prompt helpers contains the following specialists:
Art Analyst: An AI art critic that analyzes an image and provides a detailed description, critique, or artistic interpretation.
Evaluater Node / People Evaluation Node: AI judges that provide subjective evaluations based on a specific personality or set of criteria (e.g., "GLaDOS," "Attractiveness," etc.).
Keyword Extractor: An AI analyst that intelligently pulls out the most important keywords from a block of text.
QWEN Prompter: A specialized prompt writer that formats ideas specifically for the unique syntax and style of the Qwen model.
Scene Director: An AI cinematographer that takes a core scene idea and fleshes it out with directorial details.
Summary Writer: An AI editor that can combine multiple text inputs and write a coherent summary.
System Prompt: A simple but powerful utility to load a pre-written system prompt directly from a text file, keeping your workflow clean.
General Usage These nodes are almost always used as an intermediary step. The typical workflow is to take an idea (from a Master Key or a simple text box), pass it through one of the "Creative Council" nodes for refinement, and then feed the refined text into the system_prompt input of the main Gemini API node for the final generation.
Conceptual Workflow: Initial Idea -> [Creative Council Node] -> Gemini API Node -> Final Prompt
🤖 Director's Slate
A specialized AI Cinematographer that translates simple scene ideas and high-level camera directions into a complete, ready-to-use video prompt.
Philosophy & Use Case: The Director's Slate acts as your expert AI Cinematographer. It's designed for artists who want to focus on the story and mood, letting the node handle the technical prompt construction for complex camera movements. Instead of writing complex camera_schedules, you simply select a cinematic concept like "Slow Zoom" or "Fast Pan" from a dropdown menu, and the node automatically builds the correct technical prompt for you.
General Usage: The typical workflow is to write your scene description, select a motion_preset, and then feed the resulting positive_prompt_out into the system_prompt of the main Gemini API node for the final interpretation and generation.