@czottmann/pi-zai-api

Z.AI API provider extension for pi. It registers the Z.AI GLM models under a zai-api provider that points at the general Z.AI API endpoint.

Why this exists

pi already ships a built-in zai provider, but it points at the Coding Plan endpoint (https://api.z.ai/api/coding/paas/v4), which needs a Z.AI coding subscription. If you have a regular pay-as-you-go Z.AI API key, that endpoint won't work for you.

This extension registers the same GLM models under a separate zai-api provider that uses the general API endpoint (https://api.z.ai/api/paas/v4) instead, authenticated with ZAI_API_KEY. The built-in zai provider is left untouched.

Requirements

npm install -g @earendil-works/pi-coding-agent

Install

From npm:

pi install npm:@czottmann/pi-zai-api

From a local checkout:

cd path/to/pi-zai-api
npm install
pi install "$PWD"

Set up auth

Use pi's API-key flow:

pi
/login
# Choose "Use an API key", then "ZAI" (same as with the built-in `zai` provider).

Or set an environment variable before starting pi:

export ZAI_API_KEY=your-key-here

Get a key from the Z.AI Dashboard. It must be set as API keys in /login, not as a subscription.

Use

List registered models:

pi --list-models | grep zai-api

Start pi with the provider:

pi --provider zai-api

In interactive mode, /zai-api-models lists the models registered by the extension.

How it works

On startup, the extension reads pi's own built-in zai model definitions via getModels("zai") from @earendil-works/pi-ai, drops their per-model base URL, and re-registers them under the zai-api provider with the general API base URL and ZAI_API_KEY.

Everything else is carried over from pi's catalog verbatim: context window, max output tokens, reasoning support, input types, and the z.ai-specific compatibility flags (thinking format, tool streaming). Because the list is mirrored at runtime, it tracks pi's built-in zai models as you update pi: new GLM models pi adds show up here too, with the corrected endpoint.

pi resolves @earendil-works/pi-ai to its own bundled copy when it loads the extension, so the mirrored list always matches the pi version you're running.

Development

npm run check
npm run build
pi -e . --provider zai-api

A note on local dev and the mirrored list

When you load the extension from a local checkout (pi -e . or pi install "$PWD"), Node resolves @earendil-works/pi-ai to the copy in this project's node_modules, not to the one inside your pi install. If that dev copy is older than your pi, the mirrored model list reflects the dev copy and can miss models your pi actually ships.

The published package bundles no dependencies, so a real install always mirrors the pi you're running. To check the published behavior locally, install a packed build:

npm pack
mkdir -p /tmp/zai-test && tar -xzf czottmann-pi-zai-api-*.tgz -C /tmp/zai-test --strip-components=1
pi install /tmp/zai-test

Publishing

GitHub Actions publishes the package to npm when a GitHub Release is published. The release tag must match package.json exactly, with or without a leading v (v1.0.0 and 1.0.0 both work for version 1.0.0).

The workflow uses npm Trusted Publishing, so it does not need an npm token secret. Configure this package on npm with this repository and workflow file (.github/workflows/publish.yml). The workflow builds the package, runs npm run check, and publishes with npm provenance.

Author

Carlo Zottmann, carlo@zottmann.dev

• Website: https://actions.work
• GitHub: https://github.com/czottmann
• Bluesky: https://bsky.app/profile/zottmann.dev
• Mastodon: https://norden.social/@czottmann