tanstack-start（坦斯塔克开始）

TanStack Start

File-Based Routing

Routes live in src/routes/. The route tree auto-generates to src/routeTree.gen.ts - commit this file (required for type-checking).

Route Types

• __root.tsx - Root layout, wraps all routes
• index.tsx - Index route for a path segment
• demo.tsx - Layout route wrapping all demo/*.tsx children (uses <Outlet />)
• demo/*.tsx - Child routes rendered inside parent layout
• $param.tsx - Dynamic route segment

Layout Routes

Create foo.tsx alongside foo/ directory to wrap child routes:

import { createFileRoute, Outlet } from '@tanstack/react-router'

export const Route = createFileRoute('/foo')({
  component: FooLayout,
})

function FooLayout() {
  return (
    <div>
      <Outlet />
    </div>
  )
}

API Routes

Server handlers use api.*.ts naming convention:

import { createFileRoute } from '@tanstack/react-router'

export const Route = createFileRoute('/demo/api/example')({
  server: {
    handlers: {
      POST: async ({ request }) => {
        const data = await request.json()
        return Response.json({ result: data })
      },
    },
  },
})

AI Chat Integration

Client Side

Use @ai-sdk/react with DefaultChatTransport:

import { useChat } from '@ai-sdk/react'
import { DefaultChatTransport } from 'ai'

const { messages, sendMessage } = useChat({
  transport: new DefaultChatTransport({
    api: '/demo/api/chat',
  }),
})

Server Side

Use streamText from ai package with OpenRouter provider:

import { convertToModelMessages, streamText } from 'ai'
import { haiku } from '@/lib/openrouter'

const result = await streamText({
  model: haiku,
  messages: convertToModelMessages(messages),
  system: SYSTEM_PROMPT,
  tools,
})

return result.toUIMessageStreamResponse()

OpenRouter Setup

Create provider singleton in src/lib/openrouter.ts:

import { createOpenRouter } from '@openrouter/ai-sdk-provider'

export const openrouter = createOpenRouter({
  apiKey: process.env.OPENROUTER_API_KEY,
})

export const haiku = openrouter('anthropic/claude-3.5-haiku')

MCP Server

Register tools with Zod schemas in route files:

import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js'
import z from 'zod'
import { handleMcpRequest } from '@/utils/mcp-handler'

const server = new McpServer({ name: 'my-server', version: '1.0.0' })

server.registerTool(
  'toolName',
  {
    title: 'Tool Title',
    description: 'Tool description',
    inputSchema: { param: z.string().describe('Param description') },
  },
  ({ param }) => ({
    content: [{ type: 'text', text: result }],
  }),
)

export const Route = createFileRoute('/mcp')({
  server: {
    handlers: {
      POST: async ({ request }) => handleMcpRequest(request, server),
    },
  },
})

Path Aliases

@/* maps to ./src/* (configured in tsconfig.json).