pi-volcengine-provider

Pi package that registers a Volcengine Ark Coding API provider.

Install

npm package: pi-volcengine-provider

Recommended install from npm:

pi install npm:pi-volcengine-provider

To pin a specific npm version:

pi install npm:[email protected]

You can also install from GitHub:

pi install git:github.com/MacroSony/pi-volcengine-provider

For local development:

pi -e /path/to/pi-volcengine-provider

To try the published package for one run without installing it:

pi -e npm:pi-volcengine-provider

Authentication

The provider id is volcengine-plan. Pi will use ~/.pi/agent/auth.json first, then the VOLCENGINE_API_KEY environment variable as a fallback.

{
  "volcengine-plan": { "type": "api_key", "key": "your-volcengine-key" }
}

Or:

export VOLCENGINE_API_KEY=your-volcengine-key

Models

This extension follows the Volcengine Coding Plan model list and registers:

• ark-code-latest
• doubao-seed-2.0-code
• doubao-seed-2.0-pro
• doubao-seed-2.0-lite
• doubao-seed-code
• minimax-m2.7
• minimax-m3
• glm-5.2
• glm-latest
• deepseek-v4-flash
• deepseek-v4-pro
• kimi-k2.6
• kimi-k2.7-code

Run pi --list-models after installing to confirm the provider is loaded.

Thinking levels

This extension maps Pi's native thinking selector to provider-specific Chat Completions controls:

• Pi off -> thinking: { "type": "disabled" }
• Most Volcengine reasoning models: Pi low, medium, high -> thinking: { "type": "enabled" } plus matching reasoning_effort
• Pi minimal is hidden because Volcengine's reasoning_effort: "minimal" means no thinking; selecting it clamps to low
• Pi xhigh -> reasoning_effort: "max" only for models verified to accept it (ark-code-latest, deepseek-v4-flash, deepseek-v4-pro)
• glm-5.2 and glm-latest use the GLM/Z.AI thinking shape: Pi low, medium, and high send thinking: { "type": "enabled" } without reasoning_effort; Pi off sends disabled.