weckr-integration

Weckr integration

Weckr is an AI cost and margin intelligence SDK. It wraps an existing LLM client so every call is logged with its real cost, attributed per user and per feature, and optionally checked against a per user spending cap. It answers the one question a SaaS founder cannot get from a provider dashboard: which customers cost more than they pay.

Weckr sits beside the code, not in the request path. This file is the correct integration pattern. For the full API surface (every option, every error type, the provider matrix, the Python SDK) read reference.md in this skill.

1. When to use this skill

Use it whenever a user is building or already has an AI feature that calls an LLM provider and cares about any of:

• Cost visibility: what the feature actually costs to run.
• Margin per user: whether a specific customer costs more than they pay on a flat plan.
• Runaway agent cost: an agent looping and burning tokens.
• Spending limits: capping or downgrading a user who exceeds a monthly budget.

If the user only wants raw request logging with no cost or margin angle, Weckr is not the right fit and you should say so plainly.

2. How to integrate (the two line pattern)

Install:

npm install @weckr/sdk

Initialize once at boot with an api key and a plans map (plan name to monthly price in USD), then wrap each provider call with wk.chat:

import OpenAI from 'openai';
import { Weckr } from '@weckr/sdk';

const openai = new OpenAI();
const wk = new Weckr({
  apiKey: process.env.WECKR_API_KEY!,
  plans: { free: 0, pro: 29 },
});

const result = await wk.chat(openai, {
  model: 'gpt-5.4-mini',
  messages: [{ role: 'user', content: prompt }],
  userId: user.id,
  feature: 'ai-summary',
  plan: user.plan,
});

wk.chat(client, options) returns the exact result the provider would have returned, unchanged. It detects the provider from the client instance, makes the real call, and after the response resolves it sends a fire and forget log with cost and margin. The plan you pass must be a key in the constructor's plans map, or the SDK throws WeckrConfigError.