Vercel AI SDK

The Vercel AI SDK’s OpenAI and Anthropic providers both accept a baseURL. Point them at TapPass and every streamText, generateText, and tool invocation is governed.

Setup (OpenAI)

import { openai, createOpenAI } from "@ai-sdk/openai";
import { streamText } from "ai";

const tappass = createOpenAI({
  baseURL: "https://tappass.example.com/v1",
  apiKey: process.env.TAPPASS_API_KEY,
});

const { textStream } = await streamText({
  model: tappass("gpt-4o-mini"),
  prompt: "Hello",
});

for await (const chunk of textStream) process.stdout.write(chunk);

Setup (Anthropic)

import { createAnthropic } from "@ai-sdk/anthropic";
import { streamText } from "ai";

const tappass = createAnthropic({
  baseURL: "https://tappass.example.com",
  apiKey: process.env.TAPPASS_API_KEY,
});

const { textStream } = await streamText({
  model: tappass("claude-3-5-sonnet-20241022"),
  prompt: "Hello",
});

Tool calls

import { streamText, tool } from "ai";
import { z } from "zod";

const { textStream } = await streamText({
  model: tappass("gpt-4o-mini"),
  prompt: "Check the weather in Brussels",
  tools: {
    weather: tool({
      description: "Get the weather",
      parameters: z.object({ city: z.string() }),
      execute: async ({ city }) => `Sunny in ${city}`,
    }),
  },
});

Every tool call is audited — use tool_decision.rego to approve/block specific tools server-side.

Per-request headers

const tappass = createOpenAI({
  baseURL: "https://tappass.example.com/v1",
  apiKey: process.env.TAPPASS_API_KEY,
  headers: {
    "X-Agent-Id": "next-app-chat",
    "X-Governance-Flags": "pii=mask,budget=dev",
  },
});

Next.js route handler

app/api/chat/route.ts

import { streamText } from "ai";
import { tappass } from "@/lib/tappass";

export async function POST(req: Request) {
  const { messages } = await req.json();
  const result = streamText({ model: tappass("gpt-4o-mini"), messages });
  return result.toDataStreamResponse();
}