@avanor/sdk-openai

Avanor SDK — OpenAI + Anthropic auto-instrumentation.

Monkey-patches your OpenAI or Anthropic client in place so every LLM call emits an llm.call span with OTel GenAI semconv attributes (gen_ai.*).

Install

npm install @avanor/sdk @avanor/sdk-openai

Add openai or @anthropic-ai/sdk to your dependencies if they aren't already.

OpenAI quickstart

import OpenAI from 'openai';
import { Avanor } from '@avanor/sdk';
import { instrumentOpenAI } from '@avanor/sdk-openai';

Avanor.init({
  apiKey: process.env.AVANOR_API_KEY!,
  environment: process.env.NODE_ENV ?? 'dev',
  service: 'broker-portal',
});

const openai = instrumentOpenAI(new OpenAI());
// use as normal — every chat.completions.create call is instrumented
const result = await openai.chat.completions.create({
  model: 'gpt-4o',
  messages: [{ role: 'user', content: 'Hi' }],
});

Anthropic quickstart

import Anthropic from '@anthropic-ai/sdk';
import { Avanor } from '@avanor/sdk';
import { instrumentAnthropic } from '@avanor/sdk-openai';

Avanor.init({ /* ... */ });

const anthropic = instrumentAnthropic(new Anthropic());
const result = await anthropic.messages.create({
  model: 'claude-opus-4-7',
  max_tokens: 1024,
  messages: [{ role: 'user', content: 'Hi' }],
});

Attributes captured per call

| Attribute | Value | |---|---| | gen_ai.provider.name | 'openai' or 'anthropic' | | gen_ai.operation.name | 'chat.completions.create', 'messages.create', etc. | | gen_ai.request.model | The model arg (e.g., 'gpt-4o') | | gen_ai.usage.input_tokens | From the response usage field | | gen_ai.usage.output_tokens | From the response usage field | | gen_ai.response.id | The provider's response id |

Zero false-positive guarantee

The monkey-patch is OBSERVATION ONLY. The wrapped method returns the exact same value as the original — no wrapping, no transformation, no extra fields. Your downstream code does not care that the SDK is in the loop.

Docs

https://docs.avanor.ai/sdk/recipes/express-openai

License

Avanor Source License v1 (BSL-modeled, 4-year Apache 2.0 sunset) — see LICENSE.