pingcap/tidbx-kysely

TiDB Cloud + Kysely

Use this skill when a user wants to connect Kysely to TiDB Cloud (TiDB X). Default to standard TCP (Node server/runtime). Only use the TiDB Cloud serverless driver over HTTP for serverless or edge runtimes.

Workflow

1. Confirm runtime and deployment target (Node server vs serverless/edge).
2. Confirm cluster type. The serverless HTTP driver applies to Starter/Essential clusters.
3. Collect connection info (prefer a mysql:// URL in DATABASE_URL).
4. Choose the path:
   • Normal usage (default): TCP + mysql2 pool + Kysely MysqlDialect.
   • Serverless/edge: @tidbcloud/kysely dialect over HTTP.
5. Show only the matching snippet first. Include the other path only if the user asks. Use references/kysely-usage.md for full examples.

Normal usage (default)

Use this for Node servers, long-lived runtimes, or when TCP is available. This is the primary path unless the user explicitly needs serverless/edge. Uses TCP with a mysql2 pool.

import { Kysely, MysqlDialect } from 'kysely'
import { createPool } from 'mysql2'

const pool = createPool({ uri: process.env.DATABASE_URL })
const db = new Kysely({ dialect: new MysqlDialect({ pool }) })

Serverless/edge usage (HTTP)

Use this only when the runtime cannot keep TCP connections (serverless/edge). Requires the TiDB Cloud serverless driver and Starter/Essential clusters. Use from backend services only (browser origins may be blocked by CORS). See references/serverless-kysely-tutorial.md for the full walkthrough.

import { Kysely } from 'kysely'
import { TiDBCloudServerlessDialect } from '@tidbcloud/kysely'

const db = new Kysely({
  dialect: new TiDBCloudServerlessDialect({ url: process.env.DATABASE_URL }),
})

Notes

• Many users say "instance" to mean "cluster"; treat them as the same.
• Keep instructions concise; move any long docs into references.