OpenAI SDK

@lookspan/openai wraps an OpenAI client so every model call emits a Lookspan span — no OTel, no proxy, no other code changes. It returns the same client type, and calls behave identically.

Install

npm install @lookspan/openai

Usage

import OpenAI from 'openai';
import { observeOpenAI } from '@lookspan/openai';

const openai = observeOpenAI(new OpenAI());
await openai.chat.completions.create({ model: 'gpt-4o', messages });

That’s it — start npx lookspan and the call shows up as a trace with its model, token usage and (server-computed) cost.

What gets traced

The wrapper intercepts the common OpenAI surface and tags each with a span type:

Method	Span type
chat.completions.create	llm_call
responses.create	llm_call
embeddings.create	embedding

Token usage is read from the response (prompt_tokens / completion_tokens, or input_tokens / output_tokens for the Responses API). Streaming is supported: OpenAI sends usage in the final chunk (with include_usage), and the wrapper accumulates the streamed text into the span’s output.

Options

observeOpenAI(client, options) accepts the shared option bag — endpoint, agentId, sessionId, parentTraceId, captureContent, etc. For example, to attribute calls to an agent and a session:

const openai = observeOpenAI(new OpenAI(), {
  agentId: 'researcher',
  sessionId: 'run-2026-06-08',
});

Content is omitted by default. To enable Replay & judge for spans whose prompts and replies are safe to persist, opt in explicitly:

const openai = observeOpenAI(new OpenAI(), { captureContent: true });