🎨 KoboldCpp Prompt Engine for ComfyUITransform 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.&gt;&gt; to the <a target="_blank" rel="ugc" href="https://civarchive.com/articles/26503/kobold-llm-prompter-installation-and-setup-guide">Installation Guide</a> &lt;&lt;<h2 id="version-1.4-the-&quot;precision-and-control&quot;-update">Version 1.4 – The "Precision &amp; Control" Update</h2>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<ul><li>User Custom Mode:<ul><li>A new operational mode added to the <code>mode</code> selection dropdown.</li><li>This mode acts as a master override for the built-in logic presets (like SDXL or Natural Sentence).</li></ul></li><li>Dynamic System Messaging (<code>sys_msg</code>):<ul><li>A dedicated optional input field that allows you to feed your own system instructions directly into the LLM.</li><li>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.</li></ul></li><li>Multi-Phrase Filter (<code>filter_plus</code>):<ul><li>An advanced filtering input designed to handle stubborn AI artifacts.</li><li>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.</li><li>The engine automatically parses the comma-separated string, splits it by commas, and adds each entry to the regex exclusion list dynamically.</li></ul></li><li>Enhanced Debug Feedback:<ul><li>The console logging system has been upgraded for better transparency.</li><li>It now explicitly identifies when a Custom System Message is overriding the presets.</li><li>The debug log displays the active Filter+ keywords (exclusion list), so you can verify exactly which terms are being scrubbed from the final prompt.</li></ul></li></ul><h3 id="update-notes-v1.3:-the-&quot;smart-logic&quot;-overhaul"> Update Notes v1.3: The "Smart Logic" Overhaul</h3>This update significantly improves how wildcards and brackets are handled, bringing more control and variety to your prompting workflow.What's New:<ul><li>Stable Seed Logic: All wildcards and <code>{a|b}</code> 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".</li><li>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.</li><li>Recursive Wildcard Search: Wildcard files are now found automatically even within subfolders.</li><li>Enhanced Repetition Penalty: Improved the integration of the <code>repetition_penalty</code> (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.</li><li>Visual Debugging: New color-coded console logs (including <code>[POOL RESET]</code> alerts) help you track exactly how your input is being resolved. </li></ul>!!Bug Fix Update Version 1.2(1)I have refined the node's logic to fix the following issues:<ol><li>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.</li><li>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.</li></ol><hr /><h3 id="update:-version-1.2-the-&quot;smart-fusion&quot;-update">🚀 Update: Version 1.2 – The "Smart Fusion" Update</h3>This version merges my advanced <a target="_blank" rel="ugc" href="https://civarchive.com/models/2358876/smartwildcardloader">Smart Wildcard</a> logic with the high-performance Kobold LLM Prompter engine.What’s New?<ul><li>🧠 Optimized "Thinking" Mode: Specifically designed for Reasoning Models. The internal filter has been significantly improved to reliably strip <code>&lt;think&gt;</code> tags and meta-chatter, delivering a much more robust and cleaner visual prompt.</li><li>✍️ Direct Wildcard Support: You can now use wildcards (e.g., <code>__subject__</code>) directly inside the node's text input. The engine resolves them locally before sending the final context to the LLM.</li><li>♻️ Auto-Pooling &amp; 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.</li><li>📊 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., <code>Pool: 12/50</code>). </li></ul>UPDATE: Version 1.1 - The "Thinking" Update 🚀Optimized for Reasoning Models (DeepSeek-R1, Qwen-Thinking, o1).Key Features<ul><li>"Thinking" Mode: Enables Chain-of-Thought (CoT). The LLM plans the composition internally before generating the prompt.</li><li>Automatic Filtering: Removes all internal reasoning (<code>&lt;think&gt;...&lt;/think&gt;</code>) and meta-text (e.g., "Here is your prompt") so only the clean visual prompt reaches ComfyUI.</li><li>Token Buffer: Automatically adds +250 tokens in Thinking mode to prevent prompts from being cut off by lengthy reasoning.</li><li>Source Cleaner: Strips out dataset artifacts like ``.</li></ul>How to Use<ol><li>Select "Thinking" in the Mode dropdown.</li><li>Use a reasoning-capable model in KoboldCpp.</li><li>Note: <code>start_helper</code> is disabled in this mode to prioritize the <code>&lt;think&gt;</code> tag.</li><li>Enable "debug" to view the LLM's internal logic in the console.</li></ol>Tip: If the output still truncates, increase <code>max_tokens</code>. 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:<ol><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/2370862">KoboldLLMPrompter</a>: The core engine for prompt expansion.</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/2364405">Wildcard Saver</a>: Automatically archives every generated prompt.</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/2376720">LazySmartWildcards</a>: Manages dynamic inputs and wildcard processing.</li></ol><h2 id="quick-start">🚀 Quick Start</h2>    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).<h3 id="generation-modes:">🧠 Generation Modes:</h3>SDXL (Tags) * Best For: SDXL / Pony-based models.   <ul><li>Output Style: Converts input into comma-separated Danbooru-style tags</li></ul>Natural Sentence * Best For: Flux.1, SD3, or Midjourney-style prompting.<ul><li>    Output Style: Creates a cohesive, cinematic paragraph naturally fusing the subject, style, environment, and lighting.</li></ul>Z-Engineer Best For: <a target="_blank" rel="ugc" href="https://huggingface.co/BennyDaBall/Qwen3-4b-Z-Image-Engineer-V4/tree/main">Qwen3-Z-Engineer Models</a>* or similar high-parameter models.<ul><li>    Output Style: A production-focused, ~200-250 word paragraph with a deep focus on visual staging, lighting physics, and material textures.</li></ul><h3 id="key-functions:">🛠️ Key Functions:</h3>    Style Selection:<ul><li>Choose from 14 aesthetics (e.g., Cyberpunk, DSLR, Anime) or use Random to cycle styles based on the seed.</li></ul>    Start Helper:<ul><li>Force-starts the AI with specific phrases to bypass conversational "chatter" and ensure consistency.</li></ul>    Filter:<ul><li>Internal logic that automatically strips AI artifacts like "Sure! Here is your prompt" and cleans up unfinished sentences.</li></ul><h3 id="general-settings:">⚙️ General Settings:</h3>Temperature Advice:<ul><li>Use 0.2 – 0.5 for literal, prompt-loyal results.</li><li>Use 0.8 – 1.2 for creative variety and unexpected descriptions.</li></ul>Max Tokens Advice:<ul><li>        Low (50–150): Perfect for SDXL (Tags) to keep them punchy.</li><li>        High (100): Necessary for Natural Sentence or </li><li>     High+ (250+): for Z-Engineer.</li></ul>*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.

Kobold LLM Prompter

Bug Fix Update Version 1.2I have refined the node's logic to fix the following issues:<ol><li>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.</li><li>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.</li></ol>🚀 Update: Version 1.2 – The "Smart Fusion" UpdateThis version merges my advanced<a target="_blank" rel="ugc" href="https://civitai.com/models/2358876/smartwildcardloader"> Smart Wildcard </a>logic with the high-performance Kobold LLM Prompter engine.What’s New?<ul><li>🧠 Optimized "Thinking" Mode: Specifically designed for Reasoning Models. The internal filter has been significantly improved to reliably strip <code>&lt;think&gt;</code> tags and meta-chatter, delivering a much more robust and cleaner visual prompt*. </li><li>♻️ Auto-Pooling &amp; 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.</li><li>📊 Live Pool-Analytics: The console now 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., <code>Pool: 12/50</code>).</li></ul>*Successfully tested using this <a target="_blank" rel="ugc" href="https://huggingface.co/DavidAU/L3.1-Dark-Reasoning-LewdPlay-evo-Hermes-R1-Uncensored-8B">model</a> in GGUF format at Q8 quantization, as recommended.

Files with SHA256: b1a3c551c192d90275fa40746dfcf229e86a247f5ba7d5249223901e2d16de6e