Nodes/ComfyUI-ThinkingLLM/ThinkingLLM Prompt Enhancer (GGUF)
ComfyUI Node

ThinkingLLM Prompt Enhancer (GGUF)

A ComfyUI node in ThinkingLLM with 17 inputs and 2 outputs.

By goodguy1963·Created 3 months ago·Updated 25 days ago· 13
ThinkingLLM Prompt Enhancer (GGUF)
    • ENHANCED_OUTPUT
    • RAW_TRACE
    model_nameGoekdeniz-Guelmez.Josiefied-Qwen3-4B-Instruct-2507-gabliterated-v1.Q4_K_M.gguf [~2.4GB]
    prompt_text
    preset_system_prompt📝 Enhance
    custom_system_prompt
    max_tokens1024
    temperature0.70
    top_p0.90
    repetition_penalty1.10
    english_outputfalse
    deviceauto
    keep_model_loadedfalse
    seed1
    keep_last_promptfalse
    stream_tokens_to_terminalfalse
    enable_thinkingtrue
    auto_finalization_retryfalse
    hf_token
    CategoryThinkingLLM

    Inputs (17)

    NameTypeDefaultDescription
    model_nameCOMBOGoekdeniz-Guelmez.Josiefied-Qwen3-4B-Instruct-2507-gabliterated-v1.Q4_K_M.gguf [~2.4GB]GGUF text model from config or auto-detected local files. First run downloads the selected GGUF when it is not already on disk.
    prompt_textSTRINGPrompt text to enhance. Leave blank to emit the selected preset instruction as the base prompt.
    preset_system_promptCOMBO📝 EnhancePreset enhancement style. Use Custom Only when you want custom_system_prompt to fully control the instruction.
    custom_system_promptSTRINGOptional extra instruction. Required when using Custom Only; otherwise it is prepended to the selected style.
    max_tokensINT102432–16384Maximum new tokens for the enhanced prompt. Increase only when the model truncates useful detail.
    temperatureFLOAT0.700.1–1Sampling randomness. Lower is more stable; higher is more varied.
    top_pFLOAT0.900–1Nucleus sampling cutoff. Lower values restrict token choice; 0.9 is a balanced default.
    repetition_penaltyFLOAT1.100.5–2Values above 1.0 reduce repeated phrases in the enhanced prompt.
    english_outputBOOLEANfalseAsk the model to return the final enhanced prompt in English.
    deviceCOMBOautoauto prefers GPU when available. If generation is unexpectedly slow, run tools/check_llama_backend.py to verify llama.cpp GPU offload.
    keep_model_loadedBOOLEANfalseKeep the GGUF model in memory after generation so repeated prompt enhancement skips model loading.
    seedINT11–4294967295Sampling seed. Reusing it with identical inputs can reuse the saved prompt result.
    keep_last_promptBOOLEANfalseReturn the saved per-node prompt instead of generating a new one when available.
    stream_tokens_to_terminalBOOLEANfalseShow clean wrapped generated tokens in the ComfyUI terminal. Streaming bypasses fixed-seed reuse for a fresh run.
    enable_thinkingBOOLEANtrueEnable model reasoning/thinking when supported. The node still returns the cleaned final prompt, so reasoning may be hidden or empty.
    auto_finalization_retryBOOLEANfalseIf enabled, runs an extra LLM completion when the first output is empty or reasoning-only. Disabled by default so one node execution performs one generation pass.
    hf_tokenSTRINGOptional Hugging Face access token for private or gated GGUF downloads. It is passed only to the download call, never logged or cached, and the in-memory copy is dropped after the download attempt. Clear this field before saving or sharing workflows.

    Outputs (2)

    NameTypeDescription
    ENHANCED_OUTPUTSTRING
    RAW_TRACESTRING