Vercel AI SDK Integration

TekMemo provides official helpers for integrating memory into applications built with the Vercel AI SDK. These helpers live in the separate @tekbreed/tekmemo-adapter-ai-sdk adapter package and implement the framework-neutral TekMemoMemoryRuntime contract defined in core.

It allows your agents to automatically search memory for context and record new decisions or facts during a conversation.

Installation

Install the adapter alongside core and the Vercel AI SDK peer dependency:

npm install @tekbreed/tekmemo @tekbreed/tekmemo-adapter-ai-sdk ai

Core Helpers

Helper	Purpose
buildRuntimeMemoryContext()	Compiles memory-aware context (core memory, notes, recall) for system prompts.
buildRuntimeMemoryToolDefinition()	Returns an AI SDK tool definition for runtime-based memory operations.
runRuntimeMemoryTool()	Executes runtime memory tool commands with scope enforcement.
createAiSdkRuntimeFromTekmemo()	Creates an AI SDK runtime backed by a Tekmemo instance (the recommended entry point).

Example Usage

This example shows how to ground a generateText call in your project's memory. The runtime comes from a Tekmemo instance, so every operation — recall, context, writes — flows through TekMemo's hybrid engine (BM25 + fuzzy + vectors when an embedder is configured + recency + reranker).

import { generateText, stepCountIs } from "ai";
import { openai } from "@ai-sdk/openai";
import { Tekmemo } from "@tekbreed/tekmemo";
import {
  buildRuntimeMemoryContext,
  buildRuntimeMemoryToolDefinition,
  createAiSdkRuntimeFromTekmemo,
} from "@tekbreed/tekmemo-adapter-ai-sdk";

const memo = new Tekmemo({ rootDir: "./.tekmemo", projectId: "my-project" });
const runtime = createAiSdkRuntimeFromTekmemo(memo);
const access = { projectId: "my-project", userId: "user_123" };

async function chat(userPrompt: string) {
  // 1. Build a memory-aware context.
  // Reads core memory + notes and runs a hybrid recall over the query, then
  // compiles the results into a context block for the system prompt.
  const { text: system } = await buildRuntimeMemoryContext({
    runtime,
    access,
    query: userPrompt,
    baseInstructions: "You are a helpful coding assistant.",
  });

  // 2. Generate text with a memory tool.
  // The model can recall more, read core memory, and save durable notes
  // during multi-step reasoning.
  const result = await generateText({
    model: openai("gpt-4.1-mini"),
    system,
    prompt: userPrompt,
    tools: {
      memory: buildRuntimeMemoryToolDefinition({
        runtime,
        access,
        allowWrites: true,
      }),
    },
    stopWhen: stepCountIs(6),
  });

  return result.text;
}

Architectural Note

The AI SDK helpers are designed to be "pluggable." They rely on the TekMemoMemoryRuntime interface (defined in core, implemented by the adapter), and createAiSdkRuntimeFromTekmemo() adapts the full Tekmemo class to that contract — so recall, context, and writes use the same intelligent engine whether the Tekmemo instance is backed by the local filesystem (mode: "local"), a hybrid setup, or a cloud-backed client. Only the Tekmemo construction changes; the agent code is identical.