pi-provider-sicau-pmt

Pi provider for a self-hosted oh-pmt-api gateway (OpenAI-compatible), with dynamic model discovery, thinking-effort support and a /login API-key prompt

Packages

Package details

extension

Install pi-provider-sicau-pmt from npm and Pi will load the resources declared by the package manifest.

npm report

$ pi install npm:pi-provider-sicau-pmt

Package: pi-provider-sicau-pmt
Version: 0.2.0
Published: Oct 25, 2026
Downloads: 525/mo · 525/wk
Author: craigkettleman
License: MIT
Types: extension
Size: 15.4 KB
Dependencies: 0 dependencies · 1 peer

Pi manifest JSON

{
  "extensions": [
    "./src/index.ts"
  ]
}

Security note

Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.

README

pi-provider-sicau-pmt

Pi provider for a self-hosted oh-pmt-api gateway (OpenAI-compatible). Adds the gateway's models to Pi with correct context windows, output limits and thinking-effort tables, and asks for the API key through /login instead of hand-editing config files.

Install

pi install npm:pi-provider-sicau-pmt

Then log in:

/login

Pick SICAU-PMT, paste your gateway API key. It is stored in ~/.pi/agent/auth.json as an api_key credential for the sicau-pmt provider.

To try it without installing:

pi -e npm:pi-provider-sicau-pmt

What it registers

The model list comes from the gateway's /v1/models at startup. Models added to the gateway show up in Pi without editing this package; the built-in table below is only a fallback for offline startup and the first run after install.

Model	Context	Output	Thinking effort	Image input
Qwen-3.8-Flash	180K	131072	low / medium / xhigh, off	no
DeepSeek-Flash	1M	131072	low / high / max, off	no
GLM-5.3-Flash	1M	131072	low / high / max	yes
MiniMax-M3.1-Flash	1M	128000	low / medium / high / xhigh / max	yes
DeepSeek-V4.1-Flash	1M	128000	low / high / max, off	yes
Space-Bunny	1M	128000	low / medium / high / xhigh / max	yes
Step-5-Preview	1M	64000	low / medium / high	yes

The gateway's /v1/models publishes no capability metadata, so capabilities live in src/models.ts, keyed by upstream key. Context windows and output limits come from Qoder's official catalog and WorkBuddy's product config; effort tables come from the vendors' declared effort lists.

Duplicate display ids are collapsed: the gateway lists GLM-5.3-Flash and MiniMax-M3.1-Flash twice, once per upstream, and the id-form request always resolves to the same backend, so one entry is enough.

compat.supportsDeveloperRole is disabled on every model. Pi switches the system prompt to the developer role on any model with reasoning: true, and the Qoder upstream validates message roles against a serde enum that accepts only system, assistant, user and tool — a developer message rejects the whole request.

Point it at your own gateway

BASE_URL in src/index.ts is pinned to https://api.hhy.homes/v1. To use a different deployment, fork this package, change BASE_URL, and rewrite the capability table in src/models.ts for your models.

Requirements

Pi 1.4.0 or newer — the model list is refreshed through ProviderConfig.refreshModels.
A gateway that forwards reasoning_effort and reads max_completion_tokens. Gateways without that plumbing still work, but the thinking selector has no effect and the output limit falls back to the server default.

License

MIT