菜单
Set up and configure Wavelet projects. Use when the user wants to add events, queries, or subscriptions to a wavelet.config.ts, integrate the Wavelet SDK or React hooks, or set up MCP for agent integration.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/risingwavelabs/wavelet --skill wavelet命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
基于 SOC 职业分类
| name | wavelet |
| description | Set up and configure Wavelet projects. Use when the user wants to add events, queries, or subscriptions to a wavelet.config.ts, integrate the Wavelet SDK or React hooks, or set up MCP for agent integration. |
Wavelet is a reactive backend that pushes pre-computed SQL query results to apps and AI agents over WebSocket.
Everything is defined in a single config file. The full type signature:
import { defineConfig, sql } from '@risingwave/wavelet'
export default defineConfig({
database: 'postgres://root@localhost:4566/dev',
events: {
// Event ingestion. Column types: 'string' | 'int' | 'float' | 'boolean' | 'timestamp' | 'json'
game_events: {
columns: {
user_id: 'string',
action: 'string',
value: 'int',
: ,
: ,
}
}
},
: {
: {
: ,
: ,
: [, ],
}
},
: {
: sql,
: {
: sql,
: ,
},
: {
: sql,
: { : , : },
},
},
: {
: process..,
},
: {
: ,
: ,
},
})
npx wavelet init # Create wavelet.config.ts template
npx wavelet dev # Sync config to RisingWave + start server (auto-starts RisingWave)
npx wavelet push # Sync config to RisingWave without starting server
npx wavelet generate # Generate typed client at .wavelet/client.ts
npx wavelet status # Show config summary
All commands are idempotent.
After running npx wavelet generate, a typed client is available at .wavelet/client.ts.
import { useWavelet } from './.wavelet/client'
function Component() {
const { data, isLoading, error } = useWavelet('query_name')
// data is fully typed based on the query's columns
}
import { TypedWaveletClient } from './.wavelet/client'
const wavelet = new TypedWaveletClient({
url: 'http://localhost:8080',
token: 'jwt-token', // or: () => getToken()
})
// Read query
const rows = await wavelet.queries.query_name.get()
// Subscribe to live updates
const unsub = wavelet.queries.query_name.subscribe({
onData: (diff) => {
// diff.inserted, diff.updated, diff.deleted
},
})
// Write events
await wavelet.events.event_name.emit({ key: 'value' })
await wavelet.events.event_name.emitBatch([{ key: 'v1' }, { key: 'v2' }])
GET /v1/health -> { status: "ok" }
GET /v1/queries -> { queries: [...] }
GET /v1/queries/{name} -> { query: "name", rows: [...] }
GET /v1/queries/{name}?key=value -> filtered rows
GET /v1/events -> { events: [...] }
POST /v1/events/{name} -> { ok: true }
POST /v1/events/{name}/batch -> { ok: true, count: N }
WS /subscribe/{name} -> real-time diffs
Add to your MCP config (Codex Desktop, Cursor, etc.):
{
"mcpServers": {
"wavelet": {
"command": "npx",
"args": ["@risingwave/wavelet-mcp"],
"env": {
"WAVELET_DATABASE_URL": "postgres://root@localhost:4566/dev"
}
}
}
}
Available tools: list_queries, query, list_events, emit_event, emit_batch, run_sql.
Queries are standard SQL running on RisingWave. Key patterns:
-- Aggregation
SELECT user_id, SUM(amount) AS total FROM orders GROUP BY user_id
-- Windowed aggregation
SELECT user_id, COUNT(*) AS recent_orders
FROM orders WHERE created_at > NOW() - INTERVAL '1 hour'
GROUP BY user_id
-- Threshold / alert (rows appear/disappear as condition changes)
SELECT user_id, COUNT(*) AS tx_count
FROM transactions GROUP BY user_id HAVING COUNT(*) > 100
-- Join events with reference tables
SELECT o.*, u.name, p.category
FROM orders o JOIN users u ON o.user_id = u.id JOIN products p ON o.product_id = p.id
-- Window functions
SELECT symbol, price,
AVG(price) OVER (PARTITION BY symbol ORDER BY ts ROWS 50 PRECEDING) AS ma50
FROM trades
wavelet.config.ts under eventsqueries that references the event tablenpx wavelet dev (or npx wavelet push if server is already running)npx wavelet generate to update the typed clientUse the object form for the query with filterBy:
queries: {
my_query: {
query: sql`SELECT tenant_id, ... FROM ... GROUP BY tenant_id`,
filterBy: 'tenant_id',
}
}
The filterBy column must appear in the SELECT. Clients with a JWT containing { tenant_id: "t1" } only receive rows where tenant_id = 't1'.
Add a CDC source:
sources: {
prod_db: {
type: 'postgres',
connection: process.env.POSTGRES_URL,
tables: ['users', 'orders'],
}
}
Tables become available in queries as prod_db_users, prod_db_orders.