Kobold LLM Prompter

🎨 KoboldCpp Prompt Engine for ComfyUI

Transform simple ideas into high-production prompts. This node integrates local LLMs via KoboldCpp to expand basic input into descriptive Prompts, ranging from Danbooru tags to technical staging.

>> to the Installation Guide <<

Version 1.4 – The "Precision & Control" Update

With the release of Version 1.4, you now have full creative control over the engine's behavior: you can write your own Custom System Prompt and manually extend the Output Filter to suit your needs.

🛠️ New Features in v1.4

• User Custom Mode:

  • A new operational mode added to the mode selection dropdown.

  • This mode acts as a master override for the built-in logic presets (like SDXL or Natural Sentence).

• Dynamic System Messaging (sys_msg):

  • A dedicated optional input field that allows you to feed your own system instructions directly into the LLM.

  • When the "User Custom" mode is active and this field is populated, the engine uses your specific text as the primary instruction set for the generation.

• Multi-Phrase Filter (filter_plus):

  • An advanced filtering input designed to handle stubborn AI artifacts.

  • It supports comma-separated values, allowing you to strip multiple specific words, phrases, or conversational "noise" (temporarily extending the internal filter) in a single pass.

  • The engine automatically parses the comma-separated string, splits it by commas, and adds each entry to the regex exclusion list dynamically.

• Enhanced Debug Feedback:

  • The console logging system has been upgraded for better transparency.

  • It now explicitly identifies when a Custom System Message is overriding the presets.

  • The debug log displays the active Filter+ keywords (exclusion list), so you can verify exactly which terms are being scrubbed from the final prompt.

Update Notes v1.3: The "Smart Logic" Overhaul

This update significantly improves how wildcards and brackets are handled, bringing more control and variety to your prompting workflow.

What's New:

• Stable Seed Logic: All wildcards and {a|b} brackets now use MD5-hashing. This ensures that your prompts remain 100% consistent when using a fixed seed, while still being perfectly randomized on "randomize".

• Smart Auto-Pooling (Fixed): The engine now remembers used terms from your wildcard files. It will cycle through the entire list before repeating a word, ensuring maximum variety in long batches.

• Recursive Wildcard Search: Wildcard files are now found automatically even within subfolders.

• Enhanced Repetition Penalty: Improved the integration of the repetition_penalty (RepPen) for KoboldCpp. This prevents the LLM from getting stuck in loops or reusing the same descriptive adjectives too often, resulting in much more diverse and creative prompt expansions.

• Visual Debugging: New color-coded console logs (including [POOL RESET] alerts) help you track exactly how your input is being resolved.
  

!!Bug Fix Update Version 1.2(1)

I have refined the node's logic to fix the following issues:

1. Targeted Token Allocation: The +250 token bonus is now technically isolated to the "Thinking" mode only. All other modes (SDXL Tags, Natural Sentence, Z-Engineer) strictly adhere to the limit set in the UI.

2. Correct Category Recognition: A conditional logic error was corrected. The script now reliably identifies the selected mode and sends the appropriate system instructions to KoboldCpp without cross-mode interference.

🚀 Update: Version 1.2 – The "Smart Fusion" Update

This version merges my advanced Smart Wildcard logic with the high-performance Kobold LLM Prompter engine.

What’s New?

• 🧠 Optimized "Thinking" Mode: Specifically designed for Reasoning Models. The internal filter has been significantly improved to reliably strip <think> tags and meta-chatter, delivering a much more robust and cleaner visual prompt.

• ✍️ Direct Wildcard Support: You can now use wildcards (e.g., __subject__) directly inside the node's text input. The engine resolves them locally before sending the final context to the LLM.

• ♻️ Auto-Pooling & No-Repeat: Your wildcards are now handled by a smart pooling system. A file will be completely exhausted (all lines used once) before any term is repeated.

• 📊 Live Pool-Analytics: The console tracks your wildcard "health" in real-time. You can see exactly how many items are left in a file before it resets directly in the log (e.g., Pool: 12/50).
  

UPDATE: Version 1.1 - The "Thinking" Update 🚀

Optimized for Reasoning Models (DeepSeek-R1, Qwen-Thinking, o1).

Key Features

• "Thinking" Mode: Enables Chain-of-Thought (CoT). The LLM plans the composition internally before generating the prompt.

• Automatic Filtering: Removes all internal reasoning (<think>...</think>) and meta-text (e.g., "Here is your prompt") so only the clean visual prompt reaches ComfyUI.

• Token Buffer: Automatically adds +250 tokens in Thinking mode to prevent prompts from being cut off by lengthy reasoning.

• Source Cleaner: Strips out dataset artifacts like ``.

How to Use

1. Select "Thinking" in the Mode dropdown.

2. Use a reasoning-capable model in KoboldCpp.

3. Note: start_helper is disabled in this mode to prioritize the <think> tag.

4. Enable "debug" to view the LLM's internal logic in the console.

Tip: If the output still truncates, increase max_tokens. Reasoning consumes a significant portion of the context window.

Workflow Components (Included in v1.1)

This workflow automates prompt engineering by connecting your local LLM to ComfyUI. It requires the following custom nodes:

1. KoboldLLMPrompter: The core engine for prompt expansion.

2. Wildcard Saver: Automatically archives every generated prompt.

3. LazySmartWildcards: Manages dynamic inputs and wildcard processing.

🚀 Quick Start

    Installation: Save LLM_Wildcard.py in ComfyUI/custom_nodes/.

    Backend: Ensure KoboldCpp is running a compatible model (Llama, Mistral, or Qwen).

    Connection: Set the URL to your local API (default: http://127.0.0.1:5001).

🧠 Generation Modes:

SDXL (Tags) * Best For: SDXL / Pony-based models.   

• Output Style: Converts input into comma-separated Danbooru-style tags

Natural Sentence * Best For: Flux.1, SD3, or Midjourney-style prompting.

•     Output Style: Creates a cohesive, cinematic paragraph naturally fusing the subject, style, environment, and lighting.

Z-Engineer Best For: Qwen3-Z-Engineer Models* or similar high-parameter models.

•     Output Style: A production-focused, ~200-250 word paragraph with a deep focus on visual staging, lighting physics, and material textures.

🛠️ Key Functions:

    Style Selection:

• Choose from 14 aesthetics (e.g., Cyberpunk, DSLR, Anime) or use Random to cycle styles based on the seed.

    Start Helper:

• Force-starts the AI with specific phrases to bypass conversational "chatter" and ensure consistency.

    Filter:

• Internal logic that automatically strips AI artifacts like "Sure! Here is your prompt" and cleans up unfinished sentences.

⚙️ General Settings:

Temperature Advice:

• Use 0.2 – 0.5 for literal, prompt-loyal results.

• Use 0.8 – 1.2 for creative variety and unexpected descriptions.

Max Tokens Advice:

•         Low (50–150): Perfect for SDXL (Tags) to keep them punchy.

•         High (100): Necessary for Natural Sentence or 

•      High+ (250+): for Z-Engineer.

*I found this Model and the prompting template so effective that I decided to integrate them directly.


'VibeCoded' I'll try my best, or you can always ask Gemini or ChatGPT.