ComfyUI Node
Nougan LM Studio 🧠
Chat with any model loaded in LM Studio (dev mode). Supports vision, audio and video inputs.
Nougan LM Studio 🧠
- image
- audio
- video
- RESPONSE
- PROMPT
◄server_urlhttp://127.0.0.1:1234►
◄model_namelocal-model►
◄system_promptYou are a helpful assistant.►
◄generate_prompt►
◄max_tokens512►
◄temperature0.70►
◄top_p0.95►
◄streamtrue►
◄seed0►
◄image_size1024►
◄video_frames8►
◄prompt_override—►
◄embedded_image►
CategoryNougan Suite
Inputs (16)
| Name | Type | Default | Description |
|---|---|---|---|
| server_url | STRING | http://127.0.0.1:1234 | LM Studio dev-mode server. Default http://127.0.0.1:1234 |
| model_name | STRING | local-model | Use the ⟳ button on the node to auto-detect loaded models. |
| system_prompt | STRING | You are a helpful assistant. | Instructions for the assistant (system role). |
| generate_prompt | STRING | — | |
| max_tokens | INT | 512-1–65536 | -1 lets the model decide. |
| temperature | FLOAT | 0.700–2 | — |
| top_p | FLOAT | 0.950–1 | — |
| stream | BOOLEAN | true | Token-by-token streaming drives the progress bar. |
| seed | INT | 00–18446744073709550000 | — |
| image_size | COMBO | 1024 | Longest side images/frames are downscaled to before encoding. |
| video_frames | INT | 81–32 | How many evenly-spaced frames to sample from a video. |
| imageopt | IMAGE | Single image or batch — every frame becomes a vision content part. | |
| audioopt | AUDIO | Sent as an input_audio part — needs an audio-capable model. | |
| videoopt | * | IMAGE batch of frames, a VHS video dict, or a file path string. | |
| prompt_overrideopt | STRING | Wired STRING replaces the generate_prompt widget (feed the Prompt Box back in here). | |
| embedded_imageopt | STRING | Internal — populated by the on-node image drop zone. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| RESPONSE | STRING | — |
| PROMPT | STRING | — |