wrong order

[kalle07]

i have installed comfy allready...
how i run your model?
all new models worked for me: flux2-klein, ideogram, LongCat1.5, hidream_o1 ... yours not ...

wrong install order!

first install and verify the official Boogu-Image package following its README:

git clone https://github.com/boogu-project/Boogu-Image.git
cd Boogu-Image

conda create -y -n boogu python=3.10
conda activate boogu

pip install -r requirements/torch2.7-cu126.txt
pip install -e .
python utils/get_flash_attn.py
cd ..

Then install ComfyUI in the same environment:

git clone https://github.com/comfy-org/ComfyUI.git
cd ComfyUI
pip install -r requirements.txt

Install this custom node:

cd custom_nodes
git clone https://github.com/boogu-project/ComfyUI-Boogu.git
cd ..

Download the three official model checkpoints:

[RYG81]

should be native by comfyui if really want to go up to more peoples.

[Boogu-Team]

Thanks for the report.

We tried to reproduce the installation flow on our side, but the documented order works correctly in our test environment:

1. Install and verify Boogu-Image
2. Install ComfyUI dependencies in the same Python/conda environment
3. Clone ComfyUI-Boogu into ComfyUI/custom_nodes
4. Download the three model folders into ComfyUI/models/boogu

Could you share more details so we can help debug the issue?

Please include:

• OS and GPU
• Python version
• CUDA / PyTorch version
• How you installed ComfyUI
• The exact commands you ran
• The full terminal log / traceback
• Your ComfyUI/models/boogu folder structure
• Whether Boogu-Image works from its own inference scripts before launching ComfyUI

Also, ComfyUI now supports Boogu natively in the main ComfyUI repository (https://github.com/Comfy-Org/ComfyUI), so you are welcome to try the native ComfyUI integration as well.

[kalle07]

it is not native ...
i have allready installed comfy , i dont install a second time, first yours and after in same venv comfy... right?
how i install Boogu if i have already comfy?

[abzaloff]

I also don't quite understand why I need to install a separate instance of ComfyUI to run your model. I already have ComfyUI installed—how can I run your model in it?

[kalle07]

your part

RuntimeError: Could not import the boogu package. Install Boogu-Image first, for example: pip install -e /path/to/Boogu-Image.

File "...\ComfyUI\execution.py", line 542, in execute
output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "...\ComfyUI\execution.py", line 341, in get_output_data
return_values = await _async_map_node_over_list(prompt_id, unique_id, obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "...\ComfyUI\execution.py", line 315, in _async_map_node_over_list
await process_inputs(input_dict, i)

File "e:\SwarmUI\dlbackend\comfy\ComfyUI\execution.py", line 303, in process_inputs
result = f(**inputs)
^^^^^^^^^^^

File "...\ComfyUI\custom_nodes\ComfyUI-Boogu-main\nodes.py", line 158, in load
return _load_pipeline(self.KIND, self.MODEL_NAME, dtype, device, enable_model_cpu_offload, bool(force_reload))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

File "...\ComfyUI\custom_nodes\ComfyUI-Boogu-main\nodes.py", line 94, in _load_pipeline
BooguImagePipeline, BooguImageTurboPipeline = _import_boogu_pipelines()
^^^^^^^^^^^^^^^^^^^^^^^^^

File "...\ComfyUI\custom_nodes\ComfyUI-Boogu-main\nodes.py", line 55, in _import_boogu_pipelines
raise RuntimeError(
...
i can not tell your nodes where is your model, never seen such node like yours !?!

[kalle07]

native way - at least it works ... i tried several schedulers and samplers and steps always bad output

[abzaloff]

native way - at least it works ... i tried several schedulers and samplers and steps always bad output

clip type - boogu

[DreamFilmVFX]

I've tried to install and run the node with the fp8 models for my 16Gb VRAM card. but it seems that this node is made for these versions: Torch 2.7.1 + TorchAO 0.15.0
Torch 2.11.0 + TorchAO 0.17.0
I have already comfyui with its uv venv with all well-aligned dependencies. So i don't wanna break anything.
I don't know if chatgpt and claude are hallucinating, but they both say the node should be updated for Torch 2.9.1, which doesn't seem to be supported at the moment, and I have that version in my venv. If anyone has a practical way to do this without installing new instances of comfyui and without breaking existing dependencies, please let me know. The AIs also say that one possible way would be to create a second venv just for this node, but I find that cumbersome.

[VIPQualityPost]

I agree, it should be native instead of this very custom node.
I broke my comfy installation putting requirements from boogu-image in venv 😆

[Giey098]

same boogu downloaded but is not runnable on existing comfy

[Boogu-Team]

If you already have ComfyUI installed, you do not need to install a second ComfyUI environment, and you do not need to install the Boogu-Image Python package first.

Boogu is now supported natively in the main ComfyUI repository:

Comfy-Org/ComfyUI#14523

So the recommended setup is now:

1. Update your existing ComfyUI to the latest version.
2. Install/update ComfyUI requirements if needed.
3. Download the ComfyUI-repackaged Boogu model files from:

https://huggingface.co/Comfy-Org/Boogu-Image

Place them under the standard ComfyUI folders:

ComfyUI/models/diffusion_models/
ComfyUI/models/text_encoders/
ComfyUI/models/vae/
ComfyUI/models/loras/

You do not need to clone or install:

https://github.com/boogu-project/Boogu-Image

unless you want to run the standalone Boogu inference code outside ComfyUI.

The old installation instructions in this custom-node repository are confusing now that native ComfyUI support has been merged. We will update the README to make this clear.

Thanks for pointing this out.