fortytwo-pi0.1.0

Fortytwo Engine provider for Pi coding agent

alexei-ivashov Published 7 days ago MIT ~326 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.

Reports 0

No reports yet.

Readme

Fortytwo for Pi

Adds Fortytwo as a model provider to Pi. After install, Fortytwo models appear in /model next to the built-in providers, with streaming, tool calls, image input and reasoning.

Setup guide: Connect pi in the Fortytwo docs.

Requires Pi 1.4.0 or later.

Install

pi install npm:fortytwo-pi

To try it for one run without installing:

pi -e npm:fortytwo-pi

To remove it: pi remove npm:fortytwo-pi.

Add your API key

Create a key on the API keys page of the Fortytwo Platform, then use one of these:

In Pi, run /login, choose Sign in with an API key, select Fortytwo and paste the key. Pi stores it in ~/.pi/agent/auth.json.
Or set an environment variable before you start Pi:
```
export FORTYTWO_API_KEY="your-api-key"
```

Fortytwo models appear in /model only after Pi has a key for the provider.

Select the model

Run /model, search for qwen3.8 and choose qwen/qwen3.8-27b [fortytwo]. Press Ctrl+S to make it the default. From the command line:

pi --provider fortytwo --model qwen/qwen3.8-27b

While a Fortytwo model is active, the status line shows "Powered by Fortytwo".

Models

Model ID	Name	Context (tokens)	Max output (tokens)	Input	Input price (USD / 1M tokens)	Output price (USD / 1M tokens)	Cached input price (USD / 1M tokens)
`qwen/qwen3.8-27b`	Qwen3.8 27B	262,144	235,929	text, image	0.425	2.55	0.085

Current models: Models. Current prices: Pricing.

Reasoning

Use /thinking to set the reasoning level and Ctrl+S to save it. We recommend medium.

Pi level	Sent to the API as `reasoning_effort`
off	`none`: reasoning is turned off
minimal, low, medium, high, xhigh, max	the same value

Reasoning tokens count against the output limit.

Use as the default provider

Add to ~/.pi/agent/settings.json:

{
  "defaultProvider": "fortytwo",
  "defaultModel": "qwen/qwen3.8-27b",
  "defaultThinkingLevel": "medium"
}

Alternative: models.json without the package

Pi can also reach Fortytwo through ~/.pi/agent/models.json, without installing this package. The values are in the setup guide.

Environment variables

Variable	Meaning
`FORTYTWO_API_KEY`	API key, sent as the Bearer token. Not needed if you added the key with `/login`
`FORTYTWO_BASE_URL`	Overrides `https://api.fortytwo.network/v1`. For testing only

Troubleshooting

Fortytwo is not in /model. Pi has no key for the provider. Add one with /login or FORTYTWO_API_KEY, then open /model again.
Other problems. See Common problems or ask on Discord.

Development

npm install
npm run check        # typecheck and unit tests, no network

Live checks call the production API and need a key:

export FORTYTWO_API_KEY="your-api-key"
npm run test:live    # streaming, long context, reasoning, image input
npm run test:agent   # installs the package into a temporary Pi: multi-step tool task with reasoning off and medium, and a session that triggers context compaction

License

MIT