pi-wafer

A pi extension for the Wafer Pass provider — get every Wafer model in your pi model picker (DeepSeek V4 Pro, Qwen 3.5 397B, GLM 5.1, MiniMax M2.7, …), and a footer that shows your live request quota and the time remaining until your next quota window every turn.

↑12.4k ↓3.1k 47/100 1hr 18.2%/262k (auto)        DeepSeek-V4-Pro • high

No more guessing how many requests you have left or when the limit resets — pi tells you in the same status row that normally shows your spend.

Features

• Adds the wafer.ai provider to pi with all current Wafer Pass models pre-configured (context windows, reasoning support, thinking format)
• Paste-token "OAuth" flow: pi opens wafer.ai/pass, you paste your key, done
• Custom footer (only when a Wafer model is active — leaves your other providers' footers untouched) that replaces the $0.00 (sub) cost field with:
  • used/limit request quota (e.g. 47/100)
  • Time remaining in the current quota window (e.g. 1hr or 30m)
• Quota auto-refreshes on session start, model change, and after every assistant turn
• Footer ticks every minute so the countdown stays accurate without extra requests

Requirements

• pi with extension/package support
• A Wafer Pass account and API key

Install

From npm

pi install npm:@zackify/pi-wafer

Then reload pi:

/reload

Manual install

Copy the extension into your global pi extensions directory:

mkdir -p ~/.pi/agent/extensions/pi-wafer
cp index.ts ~/.pi/agent/extensions/pi-wafer/index.ts

Or install it only for one project:

mkdir -p .pi/extensions/pi-wafer
cp index.ts .pi/extensions/pi-wafer/index.ts

Then reload pi:

/reload

Usage

1. Open the model picker in pi (/model).

2. Pick any model under the wafer.ai provider.

3. The first time, pi opens wafer.ai/pass and prompts you to paste your API key. Paste it and hit enter — it's stored in ~/.pi/agent/auth.json like any other OAuth credential.

4. Use pi normally. The footer's stats row will read something like:

   ↑12.4k ↓3.1k 47/100 1hr 18.2%/262k (auto)        DeepSeek-V4-Pro • high
   

   • 47/100 — requests used / included this window
   • 1hr — time remaining until the window rolls over (30m, 5m, 0m etc.)
   • The rest is the standard pi footer (cumulative tokens, context %, model + thinking level)

If pi can't reach the quota endpoint the field shows … instead of crashing the footer.

Models

The following models are registered automatically. They live under the wafer.ai provider; pricing is reported as $0 because Wafer Pass is a flat-rate subscription.

ID	Name	Context	Reasoning
DeepSeek-V4-Pro	DeepSeek V4 Pro	256k	yes (DeepSeek thinking format)
Qwen3.5-397B-A17B	Qwen 3.5 397B	256k	yes
GLM-5.1	GLM 5.1	198k	yes
MiniMax-M2.7	MiniMax M2.7	200k	yes

To add or update models, edit the MODELS array at the top of index.ts.

Privacy notes

• Your Wafer API key is stored locally by pi in ~/.pi/agent/auth.json under the wafer.ai provider, the same place every other OAuth provider lives.
• The extension only talks to pass.wafer.ai — once on startup / model change / after each turn — to read your quota.
• No telemetry. No logs are written by this extension.

Troubleshooting

Footer shows … where the quota should be

That means the quota fetch hasn't completed (or failed silently). Common causes:

• You haven't selected a Wafer model yet — pick one with /model.
• Your API key is missing or rejected — re-run /auth wafer.ai to paste a new one.
• pass.wafer.ai is unreachable from your network.

The quota number isn't updating after a request

It refreshes on the agent_end event (i.e. once a turn completes). If you want a manual refresh, switch models away and back.

Files

• index.ts — extension source
• package.json — pi package metadata for @zackify/pi-wafer, including the pi-package keyword for the pi.dev registry
• .github/workflows/publish.yml — npm publish workflow for GitHub releases