OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Infrastructure & Serving

Hordelib

Hordelib is a Python library that wraps a pinned ComfyUI commit and exposes an API for running ComfyUI-designed inference pipelines programmatically. Distributed as horde-engine, it is the AI Horde’s default inference backend and can generate a single image from a structured payload.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Hordelib — OpenViST project profile in AI Infrastructure & Serving.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Hordelib when a Python process needs to run ComfyUI-designed image inference programmatically, or when building an AI Horde inference worker around its backend.

Who it is for

  • Python developers integrating ComfyUI-designed image workflows into applications or services.
  • AI Horde worker and backend developers using its documented inference engine.

Documented uses

  • Execute ComfyUI-designed inference pipelines from Python code.
  • Submit image-generation settings such as prompt, model, sampler, dimensions and seed through an API payload.
  • Retrieve and save an image result from a single-image inference call.
  • Provide an inference backend for AI Horde workers.
Before you choose

Requirements & limits

The source repository identifies the project as AGPL-3.0 licensed. The README names horde-engine as the PyPI package and hordelib as the Python module.

  • The README requires Python 3.12 or later.
  • The standard installation requires an NVIDIA GPU with CUDA; Linux NVIDIA installations require the CUDA Toolkit.
  • The README also describes AMD ROCm installation for Linux and a CPU build designated for testing; CPU execution is described as roughly 100 times slower.
  • Users choose a PyTorch build for their hardware. The README provides CUDA-specific package guidance and currently documents only a torch 2.9.1 ROCm 6.4 option.
  • Set AIWORKER_CACHE_HOME to a model directory before using the quick-start API flow.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Run a single-image pipeline through the Python API

Generate and save one image from a Python process using Hordelib’s documented quick-start flow.

Before you start

  • Python 3.12 or later.
  • A supported PyTorch build matched to the hardware; the standard installation requires an NVIDIA GPU with CUDA.
  • A model directory configured with AIWORKER_CACHE_HOME.

Steps

  1. Install horde-engine with the PyTorch build appropriate for the CUDA hardware.

  2. Set AIWORKER_CACHE_HOME to the model directory.

  3. Parse command-line arguments before initialization, then call hordelib.initialise() once in the process.

  4. Import HordeLib and ImageGenPayload from hordelib.api, create a payload with generation settings, and call basic_inference_single_image(payload).

  5. Save the returned result.image to a file.

Expected outcome

The documented API returns an image result that can be saved as a file.

Guide limitations

  • ComfyUI-touching imports must follow initialization; initialization clears sys.argv.
  • The standard installation instructions require NVIDIA CUDA. The README separately describes AMD ROCm on Linux and a CPU build for testing; CPU execution is described as roughly 100 times slower.
  • Audio generation is currently unsupported.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
13 October 2026
Original website title
Haidra-Org/hordelib · Official homepage title · retained in its source language
How we review sourcesSuggest a correction