pi-provider-poolside

A lean Pi provider for the Poolside Platform OpenAI-compatible API.

Features

• Discovers every model available to your account from Poolside's authenticated GET /v1/models endpoint
• Ships current Poolside Platform model metadata as an offline fallback
• Uses live context, output, modality, reasoning, and pricing metadata when Poolside returns it
• Supports Pi's thinking selector using Poolside's reasoning: { effort } request format
• Uses Pi's built-in OpenAI Chat Completions transport—no custom protocol or runtime dependency

Install

pi install npm:pi-provider-poolside

To try a local checkout:

pi -e ./src/index.ts

Authenticate

Create a key at platform.poolside.ai and export it before starting Pi:

export POOLSIDE_API_KEY="your-key"
pi

You can instead store an API-key credential for the poolside provider in Pi's auth.json or override the provider in models.json.

Select a model with /model. Refresh the remote catalog with Pi's normal model refresh/update flow (pi update --models). If discovery is temporarily unavailable, the extension retains the last successful catalog, or the documented Laguna S 2.1, Laguna M.1, and Laguna XS 2.1 fallbacks on first use. As with every authenticated Pi provider, models become selectable after a key is configured.

Thinking effort

Use Pi's thinking-level control as usual. Poolside receives OpenRouter-style reasoning configuration:

{ "reasoning": { "effort": "high" } }

Laguna M.1, Laguna XS 2.1, and dynamically discovered reasoning models expose off, minimal, low, medium, high, and xhigh. Laguna S 2.1 accurately exposes only off and max; Pi maps max to Poolside's highest wire-level value, xhigh.

Model metadata

The fallback catalog follows Poolside's supported-models documentation and model release notes:

Model	Context	Max output	Input	Reasoning
poolside/laguna-s-2.1	262,144	32,768	text	off / max
poolside/laguna-m.1	262,144	32,768	text	configurable effort
poolside/laguna-xs-2.1	262,144	32,768	text	configurable effort

An authenticated /models response replaces this list, so newly released or account-specific models appear without an extension release. Poolside's richer response fields supply metadata for those models; if a deployment returns only the basic OpenAI model shape, the extension uses conservative text-model defaults until richer metadata is available. The Laguna S 2.1 weights support up to 1M context, but the hosted Poolside Platform endpoint currently advertises a 262,144-token serving limit, which is the limit used here.

Development

npm install
npm run check

Project

• Changelog
• Contributing
• Security policy
• Release process

License

MIT