How to: integrate with the Vercel AI SDK

The TypeScript client ships a LanguageModelV1-shaped middleware that auto-injects context and persists messages around any Vercel AI SDK call. This page covers the wiring; a runnable example lives at clients/typescript/examples/vercel-ai.

What the middleware does

For every model invocation:

  1. Fetches three-tier context for the conversation (reflections, observations, recent messages) and prepends it to the prompt.

  2. Persists the user’s input message before generation.

  3. Persists the assistant’s response after generation.

If three-tier context is unavailable (bridge transport, or a service error), the middleware falls back to flat conversation history.

Basic wiring

import { generateText, experimental_wrapLanguageModel as wrapModel } from "ai";
import { openai } from "@ai-sdk/openai";
import { MemoryClient } from "@neo4j-labs/agent-memory";
import { agentMemoryMiddleware } from "@neo4j-labs/agent-memory/middleware/vercel-ai";

const memory = new MemoryClient();
const conv = await memory.shortTerm.createConversation({ userId: "alice" });

const model = wrapModel({
  model: openai("gpt-4o-mini"),
  middleware: agentMemoryMiddleware(memory, { conversationId: conv.id }),
});

const { text } = await generateText({
  model,
  messages: [{ role: "user", content: "Hello!" }],
});

Options

agentMemoryMiddleware(client, {
  conversationId: "conv-uuid",     // or a function returning one
  userId: "alice",                  // used if conversation is lazily created
  includeContext: true,             // default true on REST
  historyLimit: 20,                 // flat-history fallback cap
  persistInput: true,               // default true
  persistResponses: true,           // default true
});

conversationId can be a string or a function. The function form is useful when the id is derived from request state (e.g., from req.user.id in a Next.js route handler).

Streaming

The same middleware works with streamText because Vercel AI handles streaming via the wrapped model — the middleware’s wrapGenerate is called once the stream completes.

Edge runtimes

Pass apiKey explicitly because process.env is not available at module-init scope on Cloudflare Workers and Vercel Edge:

export default {
  async fetch(req, env) {
    const memory = new MemoryClient({ apiKey: env.MEMORY_API_KEY });
    // ... wire middleware as above
  },
};

See How-to: deploy to edge runtimes for details.