A lean Pi provider for the Poolside Platform OpenAI-compatible API.
- Discovers every model available to your account from Poolside's authenticated
GET /v1/modelsendpoint - 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
pi install npm:pi-provider-poolsideTo try a local checkout:
pi -e ./src/index.tsCreate a key at platform.poolside.ai and export it before starting Pi:
export POOLSIDE_API_KEY="your-key"
piYou 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.
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.
The fallback catalog follows Poolside's supported-models documentation and model release notes:
| Model | Context | Max output | Input | Reasoning |
|---|---|---|---|---|
poolside/laguna-s-2.1 |
1,048,576 | 131,072 | 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 keeps the fallback metadata until richer metadata is available. When /models reports a live serving limit, that value takes precedence over the fallback—for example, a deployment may advertise a smaller Laguna S 2.1 context window than the 1,048,576-token catalog value.
npm install
npm run checkMIT