pi extension
QwenCloud models
for your coding agent
One install gives pi access to Qwen3.8, DeepSeek V4, GLM-5.2, Wan image generation, and HappyHorse video — all through QwenCloud's OpenAI-compatible API.
Installation
Install as a pi extension — no config, no build step. Models load on startup.
pi install npm:pi-qwencloud-providerpi install git:github.com/jellydn/pi-qwencloud-providerexport QWENCLOUD_API_KEY="your_key_here"
Get your API key from home.qwencloud.com → API Keys. Or run pi /login and select QwenCloud — it opens the dashboard and prompts for your key.
Supported Models
Chat models, image generation, and video generation — all through one provider.
Chat & Reasoning
| Model | Reasoning |
|---|---|
| Qwen3.8 Max Preview | low · medium · high |
| Qwen3.7 Max | low · medium · high |
| Qwen3.7 Plus | low · medium · high |
| Qwen3.6 Flash | low · medium · high |
| DeepSeek V4 Pro | high · xhigh |
| GLM-5.2 | low · medium · high · xhigh |
Image & Video Generation
Wan Image Generation
/wan
Generate images from text prompts directly in pi. Supports wan2.7-image and wan2.7-image-pro models at 1K, 2K, and 4K resolutions.
HappyHorse Video
/happyhorse
Text-to-video, image-to-video, and reference-to-video generation with HappyHorse 1.1 models.
Usage
Reference models with the qw/ prefix. Switch models in-session or start with one.
# Interactive session with DeepSeek V4 Pro
pi --model qw/deepseek-v4-pro
# One-shot with high reasoning
pi --model qw/deepseek-v4-pro --thinking high -p "Design a scalable microservice architecture"
# Generate an image
pi --model qw/qwen3.7-plus -p "/wan a cyberpunk cat on a rainy street"
# List available QwenCloud models
pi --list-models qwThinking Levels
Six levels — off · minimal · low · medium · high · xhigh. Per-model support mapped to reasoning_effort.
Auth Resolution
Explicit key → QWENCLOUD_API_KEY → ~/.pi/agent/auth.json.
Dynamic Discovery
Live /models fetch at startup with 5s timeout. Falls back to curated static list on any error.
One install, all the models
Qwen3.8, DeepSeek V4, GLM-5.2, Wan, HappyHorse — all through QwenCloud's Token Plan. Static API key, OpenAI-compatible, no OAuth dance.