OpenAI & Grok / xAI

local-pii/openai wraps an OpenAI client instance structurally, so it works with the OpenAI SDK and any OpenAI-compatible provider — xAI/Grok, Groq, Together, OpenRouter, Ollama — with no dependency on the openai package.

Usage

import OpenAI from "openai"
import { withPiiOpenAI } from "local-pii/openai"

const client = withPiiOpenAI(new OpenAI({ apiKey }))

const res = await client.chat.completions.create({
  model: "gpt-5.2",
  messages: [{ role: "user", content: "email ana@acme.com" }],
})
// the provider saw a placeholder; res.choices[0].message.content has it back

Grok / xAI

Same wrapper, different baseURL:

const client = withPiiOpenAI(
  new OpenAI({ baseURL: "https://api.x.ai/v1", apiKey: process.env.XAI_API_KEY }),
)
const res = await client.chat.completions.create({ model: "grok-4", messages, tools })

What it does

• Requests — deep-anonymizes messages: content strings, tool role content (tool results), and assistant tool_calls[].function.arguments.
• Responses — rehydrates choices[].message.content and each tool_calls[].function.arguments (parse → deep-map → re-serialize).
• Streaming (stream: true) — rehydrates delta.content across chunk boundaries.

Keep one wrapped client per conversation so the session's vault stays shared across the tool loop. Defaults to the opaque token() strategy. See Tool Calls for the manual loop with createPiiChat.