@vel4ai/ai-sdk Reference

@vel4ai/ai-sdk is an optional package. It keeps @vel4ai/sdk framework-agnostic while bundling provider setup and report callbacks that AI SDK streamText() routes otherwise need repeatedly.

APIs

API	Purpose
prepareVelaStreamTextWithResponse()	Combine Vela prepare with AI SDK stream options
createVelaOpenAI()	Create an AI SDK OpenAI provider/model from a prepared execution
createVelaStreamCallbacks()	Attach onAbort / onError / onFinish to a report target
createVelaStreamTextCallbacks()	Create callbacks from a prepared execution
createVelaStreamTextOptions()	Return model plus callbacks together
createVelaStreamTextInput()	Merge app-owned input with Vela-controlled options

Usage Mapping

Pass mapUsage when AI SDK totalUsage differs from the provider-specific payload.

import { createVelaStreamCallbacks } from "@vel4ai/ai-sdk";

declare const reporter: import("@vel4ai/sdk").ExecutionResultReportTarget;

export const callbacks = createVelaStreamCallbacks<
  { error: unknown },
  { totalUsage?: null; providerUsage: { prompt: number; completion: number } }
>(reporter, {
  mapUsage(event) {
    return {
      inputTokens: event.providerUsage.prompt,
      outputTokens: event.providerUsage.completion,
    };
  },
});

Boundaries

• @vel4ai/ai-sdk depends on @ai-sdk/openai.
• It does not manage chat history, tools, system prompts, or UI state.
• Authorization and policy responses before allow belong to the core SDK.
• It uses the non-throwing reporter path so report failures do not replace model responses or provider errors.