ZeroEval TypeScript SDK

ZeroEval is an evals, A/B testing and monitoring platform for AI products.

For full documentation visit https://docs.zeroeval.com.

Features

• Span decorator & tracer API – instrument any function with a single line and capture sessions, traces and spans easily.

• Integrations – OpenAI client, Vercel AI SDK, and LangChain/LangGraph are traced automatically (optional peer deps).

• Works everywhere – Node 18+, Bun, browser (Vite / Next.js).

Feature comparison

Feature	TypeScript SDK	Python SDK
Observability	✅ Supported	✅ Supported
Datasets	⚠️ WIP	✅ Supported
Experiments	⚠️ WIP	✅ Supported

Installation

npm install zeroeval

Optional dependencies for integrations:

npm install openai
npm install ai @ai-sdk/openai
npm install langchain

Authentication

1. Environment variable (recommended)

   export ZEROEVAL_API_KEY=YOUR_KEY

2. In code

   import * as ze from "zeroeval";
   ze.init({ apiKey: "YOUR_KEY" });

The SDK auto-initialises on first span if ZEROEVAL_API_KEY is set.

Quick-start

import * as ze from "zeroeval";
import { OpenAI } from "openai";

ze.init();

const openai = ze.wrap(new OpenAI());

// Traced automatically
const completion = await openai.chat.completions.create({
  model: "gpt-4o-mini",
  messages: [{ role: "user", content: "Hello!" }],
});
console.log(completion.choices[0].message.content);

Integrations

OpenAI

import * as ze from "zeroeval";
import { OpenAI } from "openai";

ze.init();

const openai = ze.wrap(new OpenAI());

LangChain / LangGraph

import {
  ZeroEvalCallbackHandler,
  setGlobalCallbackHandler,
} from "zeroeval/langchain";

setGlobalCallbackHandler(new ZeroEvalCallbackHandler());

Vercel AI SDK

import * as ze from "zeroeval";
import { openai } from "@ai-sdk/openai";

const wrappedAI = ze.wrap(ai);

const result = await wrappedAI.generateText({
  model: openai("gpt-4o-mini"),
  prompt: "Hello, world!",
});

License

Apache 2.0