https://github.com/siliconflow/bizyair-skill

连接 bizyair.cn 云端生态，赋予 Agent 图像/视频生成、AI 应用执行与 ModelZoo 模型调用等 AIGC 能力，助力实现更高效的视觉创作与 AI 任务流自动化。
https://github.com/siliconflow/bizyair-skill

agent-skill agent-skills ai-agents aigc bizyair text-to-image

Last synced: 16 days ago
JSON representation

连接 bizyair.cn 云端生态，赋予 Agent 图像/视频生成、AI 应用执行与 ModelZoo 模型调用等 AIGC 能力，助力实现更高效的视觉创作与 AI 任务流自动化。

Host: GitHub
URL: https://github.com/siliconflow/bizyair-skill
Owner: siliconflow
License: mit
Created: 2026-05-22T03:55:47.000Z (about 2 months ago)
Default Branch: main
Last Pushed: 2026-05-22T08:44:53.000Z (about 2 months ago)
Last Synced: 2026-05-22T14:30:45.752Z (about 2 months ago)
Topics: agent-skill, agent-skills, ai-agents, aigc, bizyair, text-to-image
Language: Python
Homepage: https://bizyair.cn
Size: 108 KB
Stars: 1
Watchers: 0
Forks: 0
Open Issues: 0
Metadata Files:
- Readme: README.en.md
- License: LICENSE

Awesome Lists containing this project

README

BizyAir Skill

---

> BizyAir Skill is an AIGC skill pack designed for mainstream AI agents. Powered by BizyAir cloud compute, it removes complex setup and parameter tuning.
**With natural language alone, go from idea to finished image and video output in one smooth flow, giving your AI agent professional-grade AIGC capability.**

[Installation](#-installation) · [Usage](#-how-to-talk-to-it) · [Examples](#-how-it-interacts-with-you) · [CLI Commands](#-developer--debug-commands)

**[中文](./README.md) · [English](./README.en.md)**

****

## ✨ What it can do

---

## ⭐️ Prerequisites

- **Python 3.9+**
- A BizyAir API key ready (sign in, hover over your avatar in the top-right corner, then click ["API Keys"](https://bizyair.cn/user/api-key) to create one)
- Sufficient BizyAir balance in your account

---

## 🚀 Installation

Send this in your agent chat:

> Help me install the BizyAir skill from` https://github.com/siliconflow/bizyair-skill `

Or run directly:

```bash
gh skill install siliconflow/bizyair-skill
```

---

### Config file notes (How to configure the API key)

**Option A: use an environment variable**

```bash
export BIZYAIR_API_KEY="your BizyAir API Key"
```

**Option B: write it into the config file**

The `config.json` in the skill root stores the BizyAir API key and batch-related settings.

You can edit it manually, or just give the API key to the assistant and let it write the file for you. **(Recommended)**

```json
{
"credentials": {
"api_key": "Put your BizyAir API Key here"
},
"batch": {
"max_concurrency": 3,
"max_tasks": 5
}
}
```

> **Notes**: `max_concurrency` — maximum concurrent tasks, default 3, hard cap 5; `max_tasks` — maximum tasks in a single batch, default 5, hard cap 10.

---

## 💬 How to talk to it

Once configured, just talk to the assistant in natural language (pattern: "Use bizyair to ..."):

**🎨 Image tasks:**
* "Use bizyair to create a warm brown coffee advertising poster."
* "Use bizyair to make an elegant anime-style avatar."
* "I want to generate an image. Show me the common BizyAir menu first."

**🎬 Video tasks:**
* "Use bizyair to make a 10-second rainy neon street short video."
* "Animate this portrait image with bizyair."
* "Use bizyair to turn this portrait and this audio clip into a lip-sync video."

**🧩 AI App / Workflow tasks:**
* "Run this BizyAir app: `https://bizyair.cn/...`"
* "Help me understand what this BizyAir workflow ID does and what assets it needs."
* "Find a few BizyAir workflows suitable for e-commerce key visual retouching."

**🛠️ Status and utility tasks:**
* "Check my BizyAir API key."
* "Check my BizyAir balance."

---

## 🤖 How it interacts with you

### 1. Curated video / image menus
When your request is vague (e.g. "help me make a video"), the assistant shows a curated model menu first.

For example, you can say:
> "I want to make a video. Show me the common BizyAir video entry points first."

The assistant will respond like:
> "Pick a model 👇
>
> 【⭐ Curated · 5 common video models】
> 1. 🎬 V.3.1.Pro: cinematic mood, lens feel, stronger visual impact.
> 2. 🐎 HappyHorse: fast output, comfortable price-performance.
> 3. 🐉 Kling 3.0 Pro: stronger for action, movement, and impact, native 4K.
> 4. 🌊 Wanxiang 2.7: strong Chinese understanding, audio sync, multi-shot editing.
> 5. 💃 Seedance 2.0: character motion, dance, and role consistency.
>
> 【🧪 ModelZoo Search · ✨ NEW】
> 6. Direct calls to underlying model APIs — explicit params, predictable per-call billing.
>
> 【🔎 AI App Search】
> 7. Find ready-made workflow templates for complex multi-step pipelines.
>
> I'd suggest starting with option 4 — your request hasn't locked down a specific style yet, so Wanxiang 2.7 tends to be more stable.
> Just reply 1–7; if you already know the model name or number, we can use that directly."

### 2. Link diagnosis
If you send an unfamiliar App link, the assistant will fetch the underlying inspection result (`info`) and explain in natural language:

* What the workflow does.
* What input assets it needs (image / audio / text).
* Whether it can be run directly or requires the web interface.

For example, you can say:
> "Help me inspect this BizyAir app: https://bizyair.cn/community/app/46086"

The assistant will respond like:
> "All checked. This workflow looks healthy.
>
> - **What it does**: it's a **Wai-Illustrious anime text-to-image workflow**, with HD upscaling and hand/face repair.
> - **What you need to provide**: no base image needed, just a solid prompt.
> - **What it outputs**: a complete set of finished images (including HD upscaled versions).
> - **Execution status**: ✅ the interface is fully reachable, I can run it for you directly.
>
> So — do you want to start shaping the prompt now, or should I first lay out the size and ratio parameter card for review?"

> **Note**: comfy-ui links (`bizyair.cn/comfy-ui?id=xxx`) are not currently supported for direct execution. The assistant will prompt you to find the corresponding AI App version.

### 3. Smart parameter confirmation card
Once a model is chosen, the assistant absorbs your natural-language request and generates a **prefill confirmation card**.

For example, you can say:
> "I'll use option 4, O.2. Make a Chinese coffee poster, landscape layout."

The assistant will respond like:
> "## 🎨 O.2 Parameter Confirmation
>
> **Prompt**: Chinese coffee poster, landscape layout
> **Ratio**: 3:2 (options: 1:1 / 2:3 / 3:2 / 4:5 / 5:4 / 3:4 / 4:3 / 16:9 / ...)
> **Resolution**: 2K (options: 1K / 2K / 4K)
>
> Once confirmed, reply "run" to generate immediately, or just tell me if you want to change the ratio or resolution."

### 4. Dual-track search
Beyond the 5 curated models, the assistant can take two different search routes depending on what you need:

- **Slot 6 — ModelZoo Search**: direct calls to underlying model APIs, with explicit parameters and predictable per-call billing. Good when you already know the rough route ("I want a Kling video", "find me a specific endpoint").
- **Slot 7 — AI App Search**: BizyAir's workflow templates that bundle complex multi-step pipelines into one click. Good when you want a pre-made solution ("a workflow that swaps backgrounds", "an e-commerce key-visual template").

#### Example 1: find a ModelZoo endpoint

For example, you can say:
> "I want to call an underlying video model API directly. Search me a few kling endpoints."

The assistant will respond like:
> 🎯 **Picked a few video-side ModelZoo endpoints for you**
>
> ModelZoo gives you direct API access — explicit params, predictable per-call billing. Here's what looks closest:
>
> 1️⃣ **Kling O3.Pro · FLF to Video (channel edition)**
>
> - **Summary**: O3-architecture first/last-frame video model, supports first/last-frame constraints with smooth motion control.
> - **Category**: FLF to Video
> - **Runnable directly**: ✅ Yes, after parameter confirmation
> - **endpoint**: `kling-o3-pro-base/flf-to-video`
>
> 2️⃣ **Kling O3.Pro · text-to-video (channel edition)**
>
> - **Summary**: O3-architecture text-to-video model, sharp subjects, smooth motion, natural lighting.
> - **Category**: Text to Video
> - **Runnable directly**: ✅ Yes, after parameter confirmation
> - **endpoint**: `kling-o3-pro-base/text-to-video`
>
> Tell me a number or endpoint name and I'll move on to the parameter card.

#### Example 2: find an AI app / workflow

For example, you can say:
> "I want to generate an architectural render. Can you check what BizyAir workflows would suit it?"

The assistant will respond like:
> 🎯 **Found some BizyAir objects that fit your needs better**
>
> I filtered out the weak matches. Here are the ones worth looking at:
>
> 1️⃣ **FLUX建筑效果图高清修复**
>
> - **Best for**: Flux.1 D based route, suitable as a candidate for this round.
> - **Runnable directly**: ✅ Yes
> - **Link**: https://bizyair.cn/community/app/44479
> - **ID**: `44479`
>
> 2️⃣ **建筑效果图放大**
>
> - **Best for**: Flux.2 Klein based route, suitable as a candidate for this round.
> - **Runnable directly**: ✅ Yes
> - **Link**: https://bizyair.cn/community/app/44396
> - **ID**: `44396`
>
> Which one do you like? Just tell me the number or ID and I'll proceed.

---

## 👨‍💻 Developer & Debug Commands

**Image / Video base flow:**
```bash
# Show common image/video route menus
python3 scripts/cli.py image-menu
python3 scripts/cli.py video-menu

# Fixed search routes (AI app candidates)
python3 scripts/cli.py pick-image "" --remote
python3 scripts/cli.py pick-video "" --remote

# ModelZoo endpoint candidates (menu slot 6 / v6)
python3 scripts/cli.py pick-modelzoo-image ""
python3 scripts/cli.py pick-modelzoo-video ""

# Generate a prefilled confirmation card (fixed routes)
python3 scripts/dispatch.py --model --prefill-card --prompt ""

# General BizyAir platform search (apps / workflows / creations / MCP)
python3 scripts/cli.py search "" --remote

# Browse in-site app listings (filterable by type/sort)
python3 scripts/cli.py browse --remote-source community --modality Application --sort "Most Used"

# Multi-task / parallel — generate batch parameter cards first
python3 scripts/cli.py batch-prefill --model --batch-json ''
# batch-json format examples (each item can be a string or object):
# Short: '["promptA", "promptB", "promptC"]'
# Full: '[{"prompt":"promptA","aspect_ratio":"16:9"}, {"prompt":"promptB"}]'
# With concurrency: '{"concurrency":2,"tasks":["promptA","promptB"]}'

# Batch confirmed execution (run after prefill confirmation)
python3 scripts/cli.py batch-run --model --batch-json '' --confirm-run

# Execute a fixed route after explicit confirmation (single task)
python3 scripts/dispatch.py --model --prompt "" [other params] --confirm-run
```

**AI App / Workflow advanced flow:**
```bash
# Inspect object details and execution support from a link or ID
python3 scripts/cli.py info ""

# Generate a prefilled confirmation card for a remote app
python3 scripts/cli.py prefill "" --prompt ""

# Submit execution for a specific App
python3 scripts/cli.py run --prompt ""
```

**ModelZoo model services:**
```bash
# Search ModelZoo endpoints
python3 scripts/cli.py modelzoo-list --keyword ""

# View endpoint details and pricing
python3 scripts/cli.py modelzoo-detail
python3 scripts/cli.py modelzoo-price

# Execute a ModelZoo task
python3 scripts/cli.py modelzoo-run --param prompt=""

# Query ModelZoo async task status
python3 scripts/cli.py modelzoo-status
```

**Account utilities:**
```bash
# Validate API key and check balance
python3 scripts/cli.py check
python3 scripts/cli.py wallet
```

ecosyste.ms

Data

Tools

Indexes

Applications

Experiments

Awesome

https://github.com/siliconflow/bizyair-skill

Awesome Lists containing this project

README

BizyAir Skill