@ai-me-chat/nextjs

Next.js integration for AI-Me — route handler factory with auto-discovery, filtering, and auth forwarding.

Installation

npm install @ai-me-chat/nextjs

Quick Start

Important: Use an optional catch-all route so sub-paths (/tools, /health) are handled:

app/api/ai-me/[[...path]]/route.ts    <-- correct
app/api/ai-me/route.ts                <-- won't handle /tools or /health

Create app/api/ai-me/[[...path]]/route.ts:

import { createAIMeHandler } from "@ai-me-chat/nextjs";
import { openai } from "@ai-sdk/openai";

const handler = createAIMeHandler({
  model: openai("gpt-4o"),

  discovery: {
    mode: "filesystem",
    include: ["/api/**"],
    exclude: ["/api/ai-me/**"],
  },

  getSession: async (req) => {
    // your auth logic
    return { user: { id: "user-1", role: "admin" } };
  },

  systemPrompt: "You are a helpful AI assistant for this app.",
});

export { handler as GET, handler as POST };

Features

• Filesystem discovery — auto-scans app/api/ routes at startup, with src/app auto-detection
• OpenAPI discovery — generate tools from an OpenAPI 3.x spec (inline or remote URL)
• Route filtering — include/exclude patterns with glob support
• Auth forwarding — forwards cookies and authorization headers to your routes
• Write confirmation — POST/PUT/PATCH/DELETE require user confirmation by default
• Dynamic system prompt — inject user-specific context via a function

Endpoints

The health endpoint does NOT require authentication — use it for liveness probes and monitoring.

App Directory Detection

When using mode: "filesystem", the handler automatically locates your Next.js app directory:

1. src/app — checked first (default for npx create-next-app)
2. app — fallback for projects without a src/ layout

You can override this by setting appDir in the discovery config:

discovery: {
  mode: "filesystem",
  appDir: "src/app",       // relative to project root, or absolute
}

OpenAPI Discovery Mode

Instead of scanning the filesystem, provide an OpenAPI 3.x spec (inline or remote):

createAIMeHandler({
  discovery: {
    mode: "openapi",
    spec: {
      openapi: "3.0.3",
      info: { title: "My API", version: "1.0.0" },
      paths: {
        "/api/users": {
          get: {
            operationId: "listUsers",
            summary: "List all users",
            parameters: [
              { name: "q", in: "query", schema: { type: "string" }, description: "Search by name" }
            ],
            responses: { "200": { description: "User list" } }
          }
        }
      }
    }
  },
  ...
})

Or fetch from a remote URL:

discovery: {
  mode: "openapi",
  specUrl: "http://localhost:3000/api/openapi.json"
}

When to use OpenAPI mode:

• Your app uses src/app and filesystem detection fails
• You want custom tool names via operationId
• You want rich parameter descriptions for better AI understanding
• You want to expose only specific endpoints (include/exclude still work)

Dynamic System Prompt

Inject user-specific context by passing a function:

createAIMeHandler({
  systemPrompt: async (session) => {
    const settings = await db.settings.findUnique({ where: { userId: session.user.id } });
    return `You are an assistant for ${settings.companyName}. The user is ${session.user.name}.`;
  },
})

Peer Dependencies

Documentation

Full setup guide and API reference: github.com/aselims/ai-me-chat

License

MIT