@typochat-sdk/core

Proxy-aware fetch wrapper for routing AI API requests through a centralized proxy. Strips provider API keys from the client and authenticates via your proxy token.

Installation

npm install @typochat-sdk/core

Usage

import { createProxyFetch } from "@typochat-sdk/core";

const proxyFetch = createProxyFetch({
  proxyUrl: "https://your-proxy.example.com",
  apiKey: "tmk_xxxxx",
});

The returned proxyFetch is a drop-in replacement for fetch. Pass it to any AI SDK that accepts a custom fetch.

Google

import { createGoogleGenerativeAI } from "@ai-sdk/google";
import { createProxyFetch } from "@typochat-sdk/core";

const proxyFetch = createProxyFetch({
  proxyUrl: "https://your-proxy.example.com",
  apiKey: "tmk_xxxxx",
});

const google = createGoogleGenerativeAI({
  apiKey: "proxy-managed",
  fetch: proxyFetch,
});

const model = google("gemini-3.1-flash-lite-preview");

Google Vertex

import { createVertex } from "@ai-sdk/google-vertex/edge";
import { createProxyFetch } from "@typochat-sdk/core";

const proxyFetch = createProxyFetch({
  proxyUrl: "https://your-proxy.example.com",
  apiKey: "tmk_xxxxx", // express mode
});

const vertex = createVertex({
  apiKey: "proxy-managed",
  fetch: proxyFetch,
});

const model = vertex("gemini-3.1-flash-lite-preview");

What it does

• Rewrites request URLs to route through your proxy
• Strips sensitive headers (Authorization, x-goog-api-key) so provider keys never leave the server
• Adds proxy authentication automatically
• Supports streaming via duplex: "half"

API

createProxyFetch(config): typeof fetch

| Parameter | Type | Description | |-----------|------|-------------| | config.proxyUrl | string | Base URL of your proxy | | config.apiKey | string | Proxy authentication token |

getBaseUrl(proxyUrl?): string

Resolves the proxy base URL.

License

MIT