@gsengai/openai

Drop-in wrapper for the OpenAI Node SDK that logs a hash-only evidence record per text generation — chat.completions.create and responses.create, streaming and non-streaming. Supports compliance with EU AI Act Article 50 and California SB 942.

Part of gsengai, open-source transparency and provenance tooling for AI product teams.

Install

pnpm add @gsengai/openai @gsengai/core

Requires Node ≥ 22.

Usage

import OpenAI from "openai";
import { createEvidenceStore } from "@gsengai/core";
import { withEvidence } from "@gsengai/openai";

const store = createEvidenceStore({ path: "./evidence.db" });
const client = withEvidence(new OpenAI(), { store, systemId: "my-feature" });

// Use `client` exactly as before — every text generation is fingerprinted
// (SHA-256, raw + normalized) and appended to the tamper-evident evidence log.

The wrapper never mutates or blocks model responses. It hashes the generated text in memory, appends a hash-only evidence record, and returns the SDK's response unmodified. If the store fails, the default is fail-open: the response is still returned, a loud warning is logged, and getLostRecordCount() increments — failMode: 'strict' opts into throwing instead.

Documentation

• Quickstart — runs keyless against a mock client
• Capture coverage — read this before relying on the evidence log (n > 1, Stream.tee(), .withResponse() semantics)

Positioning

This package supports compliance with EU AI Act Article 50 and California SB 942. It does not make you compliant — compliance depends on your system, your deployment context, and your processes. It is not legal advice.

License

Apache-2.0