Rabbit Hole

Enterprise Custom Image Generation Pipeline
Inspired by ComfyUI, Reimagined for Production

---

Why rabbit_hole?

rabbit_hole was created to bring powerful, customizable image generation pipelines into enterprise and research environments.
While ComfyUI is highly visual and flexible, it has limitations for production use:

• Node development is complex and not easily extensible.
• Workflow customization requires internal code changes.
• Integration with external Python libraries is difficult.
• Maintaining reproducibility, batch operations, and versioned config is not straightforward.

rabbit_hole was built to solve these, enabling rapid, modular, and automated workflows—fully in Python and easy to extend.

---

⭐ Update

• Jun 09, 2025: Added Flux T2I support (tested). Flux ControlNet, IPAdapter, and Inpainting support coming soon.

• Jun 05, 2025: Added IPAdapter-Plus (ViT-H + SDXL support, tested). Initial validation in progress.

---

Features

• Composable Pipeline Building
  Build custom flows as Python classes by connecting modular tunnel steps.
• YAML Configuration
  Use static, version-controlled YAML files for models, runtime, and batch settings.
• Batch & Automation Ready
  CLI and API friendly—ideal for MLOps, research, and cloud-scale workflows.
• Model Agnostic & Extensible
  Easily swap or extend with custom models, samplers, or encoders.
• Production Focus
  Robust logging, memory management, and CUDA setup for reliability.

---

Directory Structure

rabbit_hole/
├── assets/ # Project assets (e.g. logo, example images)
├── configs/ # YAML config files (static settings, models)
├── cores/ # Core utility modules (ComfyUI-adapted, may be replaced)
├── defaults/ # Logging, CUDA, and default environment setup
├── examples/ # Example results
├── executors/ # Custom flow classes—your inference logic
├── models/ # Model weights/checkpoints by type
├── tools/ # Utilities for image/latent/tensor ops
├── tunnels/ # Modular, reusable pipeline steps
├── inference.py # Main pipeline entry point
├── pyproject.toml
├── README.md
├── requirements.txt

---

Quick Start

Tested Environment: CUDA 12.4, Python 3.12, torch 2.6.0+cu124

# 1. Python environment setup
uv init --bare --python 3.12
uv sync --python 3.12

# 2. Activate venv
source .venv/bin/activate      # Linux & Mac
.venv\Scripts\activate        # Windows

# 3. Install requirements
uv pip install -r requirements.txt

# 4. (Recommended) Install CUDA/PyTorch stack
uv pip install torch==2.6.0+cu124 torchvision==0.21.0+cu124 torchaudio==2.6.0+cu124 xformers==0.0.29.post3 --index-url https://download.pytorch.org/whl/cu124
uv pip install triton-windows==3.2.0.post18   # (Windows only)
uv pip install triton==3.2.0  # (Linux/Mac)

Recommended Inference Launch

python -m inference --fast --use-sage-attention
# or
python3 -m inference --fast --use-sage-attention

Tune options and config path as needed.
See configs/README.md for YAML configuration.

Example: Build a Flow

from executors.example_text_to_image import T2IExecutor

executor = T2IExecutor()
executor(
        "1girl, izayoi sakuya, touhou, solo, maid headdress, maid, apron, short sleeves, dress, closed mouth, white apron, looking at viewer, maid apron, puffy short sleeves, frills, puffy sleeves, black dress, frilled apron, hair between eyes, outdoors, blush, throwing knives, knives, serious face, upper body, masterpiece, best quality, very aesthetic, absurdres",
        "wings, nsfw, low quality, worst quality, normal quality,",
        1024,
        1024,
        20,
        7.0,
        1241234454,
    )

Result Examples

Executor	Result
T2IExecutor
T2IHireFixExecutor
T2ILoRAExecutor
IT2IExecutor
IPAdapterIT2I

Contributing

We welcome contributions!
To get started:

1. Fork this repository.
2. Create a new branch for your feature, fix, or improvement.
3. Make your changes (add a Tunnel, Tool, bug fix, etc).
4. Open a Pull Request (PR) to the main branch.
5. Your PR will be reviewed and, once approved, merged.
6. You’ll be listed as a project contributor!

For detailed guidelines, see CONTRIBUTING.md.

ToDo

• Publish core inference code
• Add FastAPI + Queue based asynchronous generation pipeline
• Implement LLM and external API tunnels
• Publish a paper

References

• ComfyUI GitHub
• ComfyUI Documentation

Donate...