comfyui_qwen3vl_uncensored

ComfyUI Qwen3-VL Prompt Generator (Uncensored)

A ComfyUI custom node that generates/refines prompts using a local Qwen3-VL/llama.cpp backend (CLI or llama-server), with editable system prompts and per-preset max_tokens via JSON.

Features

• Prompt enhancement presets (system prompts stored in system_prompts.json)
• Per-preset max_tokens defaults (also from JSON)
• Works offline with local inference
• CLI fallback if llama-server is not reachable
• Output cleanup (spinner/control chars)

Install (manual)

1. Copy/clone into your ComfyUI custom_nodes/ folder:
   • Folder name: comfyui_qwen3vl_uncensored
2. Restart ComfyUI.

Configure presets / personality

Edit:

• system_prompts.json

You can define:

• _preset_prompts (order shown in UI)
• qwen_text.styles.<preset>.system_prompt
• qwen_text.styles.<preset>.max_tokens
• defaults.max_tokens

Security / Privacy

• No API keys required.
• Runs locally.
• Check code before use (recommended).

Troubleshooting

• If you see llama-server not reachable, the node will fall back to CLI automatically.
• If a JSON edit breaks parsing, restore from the .bak_*.json backup or fix syntax.

License

MIT (see LICENSE).