🎨 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

Update Notes v1.3: The "Smart Logic" OverhaulThis 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 (No-Repeat): 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>

Files with SHA256: e4062af7d70c912857bc41a65fc7f71cae26a4f27f54334ee3c849b061e4ca96