@qubu/adapter-aws-rds-data-api

Experimental: this package is not production-ready. It has no provider-backed CI or live Aurora validation. Production hardening needs funded AWS infrastructure and an external maintainer.

Choose the dialect-specific entry point for the Aurora engine you use. Both entry points adapt the AWS SDK's RDSDataClient to Qubu, and neither requires an engine option:

import { RDSDataClient } from "@aws-sdk/client-rds-data"
import { rdsDataApiAdapter } from "@qubu/adapter-aws-rds-data-api/postgres"
import { qubu } from "qubu"

const client = new RDSDataClient({})
const db = qubu(
  rdsDataApiAdapter(client, {
    resourceArn: process.env.AURORA_RESOURCE_ARN!,
    secretArn: process.env.AURORA_SECRET_ARN!,
    database: "app",
  }),
)

For Aurora MySQL, change only the import:

import { rdsDataApiAdapter } from "@qubu/adapter-aws-rds-data-api/mysql"

Each adapter's dialect preserves its PostgreSQL or MySQL policy while rendering AWS named placeholders (:p1, :p2, ...). The adapter encodes nulls, booleans, numbers, strings, dates, bytes, and JSON as Data API fields; it maps result metadata into object rows, affected-row counts, and generated insert IDs. Callback transactions use the Data API's begin/execute/commit or rollback transaction-ID sequence.

Limitations

• No streaming, migration adapter, or nested transaction/savepoint API is provided.
• AWS errors pass through unchanged. Abort signals are forwarded to AWS SDK requests, but server-side cancellation guarantees have not been validated.
• This integration needs funded Aurora test infrastructure or an external maintainer before production use; provider-backed production behavior is not claimed.
• Data API decimal and long results default to string form so precision is not silently lost; override resultSetOptions only when numeric conversion is intentional.
• The schema option is rejected when executing statements; qualify schema identifiers in SQL instead.
• Non-finite numbers and unsafe integer numbers are rejected. Pass large integers as bigint or strings; dates must be valid and JSON values must be serializable.