omp packages

pi-provider-volcengine-ark0.4.0

pi provider extension for Volcengine Ark (火山方舟, OpenAI-compatible coding endpoint)

ymy88 Published 3 days ago MIT ~386 downloads/mo

Untested on omp

This is a pi package. No one has reported whether it works on omp yet. If you try it, add your result below.

Did it work for you?

Result

One report per person. A new report replaces your old one.

Reports 0

No reports yet.

Readme

pi-provider-volcengine-ark

A pi provider extension for Volcengine Ark (火山方舟, OpenAI-compatible coding endpoint).

Setup

Set your Ark API key as an environment variable:

export ARK_API_KEY=...

Install

pi install npm:pi-provider-volcengine-ark

Then in pi: run /reload (or restart pi), then pick a model with /model → volcengine-ark/<model>.

Endpoint

https://ark.cn-beijing.volces.com/api/coding/v3 — the public Volcengine Ark coding endpoint.

Available models

Model ID Input Context Output Notes
doubao-seed-2.1-pro text+image 1M 262k thinking toggle + effort
doubao-seed-2.1-lite text+image 1M 262k thinking toggle + effort
doubao-seed-2.0-mini text+image 262k 131k thinking toggle + effort
doubao-seed-evolving text+image 1M 262k thinking toggle + effort, weekly-updated
deepseek-v4.1-flash text+image 1M 384k thinking toggle, effort high
kimi-k3 text+image 1M 131k thinking toggle + effort
kimi-k2.7-code text+image 262k 32k always thinks, reasoning captured
glm-5.3 text 1M 128k reasoning_effort low..max (low ≈ off)
glm-5.3-flash text+image 1M 131k reasoning_effort low..max (low ≈ off)
minimax-m3 text+image 512k 131k always thinks, reasoning captured

All model facts were verified against the live endpoint (2026-09): max_tokens caps via the endpoint's InvalidParameter error, context windows and input modalities via GET /models plus image probes, and thinking parameters via behavior diffs. glm-5.2, kimi-k2.6, and minimax-m2.7 were delisted from the coding endpoint and removed.

Doubao, kimi-k3 and deepseek models use Ark's thinking: {"type": "enabled"|"disabled"} toggle — off genuinely disables thinking. kimi-k2.7-code no longer honors enable_thinking (thinking stays on), so it's configured as always-thinking with reasoning captured. Per-model max_tokens caps: glm 128000/131072, deepseek-v4* 393216, minimax 131072, kimi 32768/131072, doubao 131072/262144.

How it works

pi.registerProvider("volcengine-ark", { ... }) registers the provider with its endpoint, auth ($ARK_API_KEY env var), and a per-model catalog (context window, output limit, pricing, reasoning/compat options). The extension only runs inside pi — list it as a peerDependency on @earendil-works/pi-coding-agent.

License

MIT