@cirthan/pi-cirthan-provider
Pi provider package for the Cirthan API.
Package details
Install @cirthan/pi-cirthan-provider from npm and Pi will load the resources declared by the package manifest.
$ pi install npm:@cirthan/pi-cirthan-provider- Package
@cirthan/pi-cirthan-provider- Version
0.9.2- Published
- Aug 2, 2026
- Downloads
- 438/mo · 358/wk
- Author
- qmx
- License
- MIT
- Types
- extension
- Size
- 11.6 KB
- Dependencies
- 0 dependencies · 2 peers
Pi manifest JSON
{
"extensions": [
"./pi-cirthan-provider.ts"
]
}Security note
Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.
README
pi-cirthan-provider
Cirthan model provider for pi. Registers the cirthan provider using Cirthan's OpenAI-compatible API.
Models
Models are discovered from Cirthan's /v1/models endpoint and cached by pi. The bundled fallback includes:
qwen3.6-35b-a3bqwen3.6-27bdeepseek-v4-flash-0731
Both Qwen 3.6 models support off, low, medium, high, and xhigh reasoning efforts.
DeepSeek V4 Flash 0731 supports off, low, and high. medium, xhigh, and max are
disabled because the served context window is limited to 180,000 tokens and the native
model exposes only low/high/max effort levels.
Setup
Provide an API key via environment variable or ~/.pi/agent/auth.json.
Option 1: Environment variable
export CIRTHAN_API_KEY="..."
# optional
export CIRTHAN_BASE_URL="https://api.cirthan.com/v1"
Option 2: auth.json (recommended)
Add to ~/.pi/agent/auth.json:
{
"cirthan": {
"type": "api_key",
"key": "your_key_here"
}
}
Usage
- List/switch models:
pi /model - Use the recommended model:
pi --model cirthan/qwen3.6-35b-a3b - Use the other Qwen model:
pi --model cirthan/qwen3.6-27b