bizytrd

bizytrd gives ComfyUI ready-to-use nodes for third-party generative models — video, image, audio, and text — served through the BizyAir platform. It also ships a standalone Python SDK (bizytrd_sdk) for calling the same models outside ComfyUI.

Wire a node into your graph, type a prompt, and get results back as native ComfyUI outputs. Media upload, task polling, and result conversion are handled for you.

Installation

pip install bizytrd

Requires Python 3.10+.

If you use ComfyUI, you normally get bizytrd through a BizyAir host plugin (for example BizyAir-CloudBerry or BizyAirCloudPlugin): their requirements.txt pulls in bizytrd automatically when the plugin is installed.

Note: importing bizytrd itself requires a ComfyUI runtime. To call the models from plain Python scripts or services, use the SDK instead — it has no ComfyUI dependency.

Configuration

Nodes authenticate against the BizyAir platform with a single API key — no per-node settings.

Pick one of the following (in order of recommendation):

1. Environment variable:

   export BIZYAIR_API_KEY=sk-your-key
   

2. An api_key.ini file on the search path (for example ~/.bizytrd/api_key.ini or next to your ComfyUI install):

   [auth]
   api_key = sk-your-key
   

3. Existing host-plugin config ~/.BizyAirPlus/apikey.ini is picked up automatically, so users of the BizyAirPlus plugin do not need to configure anything twice.

Advanced overrides (base URLs, timeouts, polling intervals) are listed in the SDK guide.

Supported models

The node catalog is generated from the BizyAir model catalog, so new models appear as they are enabled on the platform — no code change and no new release needed. Representative families currently available:

• Video — Wan 2.7, Seedance 2.0 (incl. mini / fast), Google Veo 3.1 (fast / lite / pro), Kling (o3 / 3.0), Vidu Q3 (pro / turbo), Grok Imagine, Skyreels v4, DreamActor 2.0, HappyHorse 1.0, Hailuo 2.3
• Image — GPT-Image 2, Qwen-Image 2.0 / Pro, Nano-Banana (2 / Pro), Seedream 4.5 / 5.0, Grok Image, Flux Klein, Ideogram 4, Z-Image / Z-Image Turbo, SeedVR2 upscale, and image-editing utilities (Birefnet, Kontext-Dev-LoRA)
• Music & audio — Mureka (v7.6 – v9, incl. song generation, lyrics, BGM), Ace-Step, Qwen3-TTS custom voice
• Text & vision LLM — Gemini 3.5 / 3.6 Flash (LLM / VLM), SiliconFlow chat & vision nodes, Joycaption3 captioning
• 3D — Tripo3D H3.1, Pixal3D

Each node appears in ComfyUI under the BizyTRD category, with localized (Chinese) names and tooltips when ComfyUI runs in a Chinese locale.

Python SDK quickstart

import asyncio

from bizytrd_sdk import AsyncBizyTRD


async def main():
    async with AsyncBizyTRD(api_key="sk-your-key") as client:
        task = await client.create_task(
            "wan2.7-image",
            {"model": "wan2.7-image-pro", "prompt": "a cinematic mountain lake at sunrise"},
        )
        result = await client.wait_for_task(task.request_id)
        downloaded = await client.download_outputs(result.outputs)
        print("urls:", downloaded.urls)


asyncio.run(main())

A matching sync client (BizyTRD) covers uploads and everything the async client does not. Full usage: SDK guide.

Using bizytrd from your own plugin

from pathlib import Path

from bizytrd import get_node_mappings, sync_web_assets

NODE_CLASS_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS = get_node_mappings()
sync_web_assets(Path(__file__).resolve().parent / "web")

This merges all BizyTRD nodes into your plugin's mappings and syncs the bundled frontend assets (price badges, auto-grow inputs, localization) into your plugin's web directory. See the development guide for the full integration API.

Project & docs

• SDK guide — complete SDK usage (clients, config, error types)
• Development guide — architecture, repository layout, release pipeline, roadmap (for contributors and integrators)
• AGENTS.md — engineering commands: environment setup, tests, lint, CI

Source: https://github.com/siliconflow/bizytrd. Releases are published to https://pypi.org/project/bizytrd/.