ComfyUI Node
💬 AITEC LLM Chat
A ComfyUI node in AITEC/LocalLLM with 11 inputs and 3 outputs.
💬 AITEC LLM Chat
- model
- text
- used_model
- status
◄system_promptYou are a helpful assistant.►
◄promptWrite a creative image generation prompt.►
◄temperature0.70►
◄top_p0.95►
◄max_tokens4096►
◄remove_thinktrue►
◄remove_chatmltrue►
◄suppress_thinkingfalse►
◄reset_kv_cachetrue►
◄unload_after_runfalse►
CategoryAITEC/LocalLLM
Inputs (11)
| Name | Type | Default | Description |
|---|---|---|---|
| model | LLM_MODEL | — | |
| system_prompt | STRING | You are a helpful assistant. | — |
| prompt | STRING | Write a creative image generation prompt. | — |
| temperature | FLOAT | 0.700–2 | — |
| top_p | FLOAT | 0.950–1 | — |
| max_tokens | INT | 409664–32768 | Thinking model with 4096 or more is recommended |
| remove_think | BOOLEAN | true | Remove the <think>...</think> block (Qwen3, Gemma4, etc.) |
| remove_chatml | BOOLEAN | true | Remove repetitions of <|im_start|>assistant～<|im_end|> and return only the first response (Gemma4, etc.) |
| suppress_thinking | BOOLEAN | false | When enabled, adds an inference suppression instruction to the system prompt (for Thinking models such as Qwen3 and Gemma4) |
| reset_kv_cache | BOOLEAN | true | Reset the KV cache before inference. Set to ON to prevent context exhaustion (recommended for reasoning models such as Qwen3). Set to OFF to retain conversation history. |
| unload_after_run | BOOLEAN | false | Unload the model after execution to free up VRAM. You will need to reload it when you run it again. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| text | STRING | — |
| used_model | STRING | — |
| status | STRING | — |