ComfyUI Node
Allma Generate
A ComfyUI node in Allma/llm with 11 inputs and 4 outputs.
Allma Generate
- connectivity
- model
- images
- audios
- output_prompt
- thinking
- assembled_system_prompt
- status
◄use_image_metadatatrue►
◄read_lora_metadatatrue►
◄preset▾►
◄user_prompt►
◄system_prompt►
◄enabledtrue►
◄duration—►
CategoryAllma/llm
Inputs (11)
| Name | Type | Default | Description |
|---|---|---|---|
| connectivity | ALLMA_CONNECTIVITY | — | |
| use_image_metadata | BOOLEAN | true | When ON, each connected image is traced back through the graph to the file it was loaded from, and that file's embedded prompt metadata is added to the system prompt — so the model can reason over the original prompt/model/LoRAs behind the picture. |
| read_lora_metadata | BOOLEAN | true | ON: when MODEL is connected, inject each LoRA's full sidecar block (trigger_words + notes + usage_tips + author description) into the system prompt so the LLM can mine the author's prompt-format guidance. Costs more tokens + more reasoning effort. OFF: still inject trigger_words (they are literal tokens the LoRA needs to activate and are practically free), but skip the rest. |
| preset | COMBO | 1 options: (none) | |
| user_prompt | STRING | — | |
| system_prompt | STRING | — | |
| enabled | BOOLEAN | true | OFF: skip the LLM entirely and pass user_prompt straight through to 'output_prompt'. Nothing is sent to Allma, so the graph still runs with the backend down. ON: normal enhancement. |
| modelopt | MODEL | — | |
| imagesopt | COMFY_AUTOGROW_V3 | — | |
| audiosopt | COMFY_AUTOGROW_V3 | — | |
| durationopt | FLOAT | Clip length in SECONDS — wire the same FLOAT that drives the workflow's length. Stated in the system prompt so the model scripts the action to fit the real clip instead of guessing. |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| output_prompt | STRING | — |
| thinking | STRING | — |
| assembled_system_prompt | STRING | — |
| status | STRING | — |