warmwinter (TypeScript / JavaScript SDK)

The smallest wrapper around an AI/agent call that asks the one question the Warm Winter gate exists to answer: is this cheap answer trustworthy enough to act on — or should we escalate, or abstain? You keep executing; we only judge. Zero runtime dependencies (built-in fetch; Node 18+ or any browser/edge runtime).

npm install warmwinter

import { WarmWinter } from "warmwinter";

const ww = new WarmWinter({ apiKey: "ww_..." });   // mint at /api/v1/gate/keys

const answer = await ww.guard({
  domain: "compute", decisionType: "model_route", statedConfidence: 0.82,
  cheap: () => smallModel(prompt),
  escalate: () => bigModel(prompt),
  verify: (out) => out != null,
});

Every decision prints a prompt-blind verdict line:

[warmwinter] ACT | compute/model_route | verified | medium stakes

Only decision coordinates and gate metadata are logged—never prompts, context, candidate actions, or outputs. Pass logVerdicts: false to silence it, or logger: (line) => appLogger.info(line) to use your application logger.

Or drive the halves yourself with decide(...) / outcome(...).

Seamless alternative: point your existing OpenAI/Anthropic client's baseURL at the gateway and change nothing else — see https://warmwinter.io/gate.