Vercel AI SDK

In AI SDK routes, @vel4ai/ai-sdk bundles createOpenAI(), model, onFinish, onError, onAbort, and reporter sharing.

Shortest Route

prepareVelaStreamTextWithResponse() returns either a non-allow response or an allowed result with streamTextOptions for AI SDK.

import { prepareVelaStreamTextWithResponse } from "@vel4ai/ai-sdk";
import { createVelaIntegrationFromEnvironment } from "@vel4ai/sdk";
import { streamText } from "ai";

declare const process: { env: Record<string, string | undefined> };
declare const messages: unknown[];

export async function runStream() {
  const vela = createVelaIntegrationFromEnvironment({ env: process.env });
  const prepared = await prepareVelaStreamTextWithResponse(
    vela,
    { estimatedCostUsd: 0.05 },
    {
      onReportError(error) {
        console.error("Vela report failed", error);
      },
    },
  );

  if (prepared.outcome !== "allow") {
    return prepared.response;
  }

  return streamText(prepared.createStreamTextInput({ messages }));
}

Allow-Only Dynamic Import

In routes that should avoid loading the AI SDK runtime in setup or deny paths, prepare with the core SDK first, then import @vel4ai/ai-sdk and ai only after allow.

import { createVelaIntegrationFromEnvironment } from "@vel4ai/sdk";

declare const process: { env: Record<string, string | undefined> };
declare const messages: unknown[];

export async function runAllowOnlyStream() {
  const vela = createVelaIntegrationFromEnvironment({ env: process.env });
  const prepared = await vela.prepareExecutionWithResponse({
    estimatedCostUsd: 0.05,
  });

  if (prepared.outcome !== "allow") {
    return prepared.response;
  }

  const [{ createVelaStreamTextInput }, { streamText }] = await Promise.all([
    import("@vel4ai/ai-sdk"),
    import("ai"),
  ]);
  const reporter = prepared.createResultReporter();

  return streamText(
    createVelaStreamTextInput(prepared, { messages }, { reporter }),
  );
}

Low-Level Helpers

Helper	Purpose
createVelaOpenAI()	Create an AI SDK OpenAI provider/model from a prepared execution
createVelaStreamCallbacks()	Attach AI SDK callbacks to an existing report target
createVelaStreamTextCallbacks()	Create callbacks from a prepared execution
createVelaStreamTextOptions()	Get model and callbacks together
createVelaStreamTextInput()	Merge app-owned streamText() fields with Vela-controlled options