SDK Reference

@getflect/sdk is the TypeScript client for accessing Flect-managed resources. Its whole surface is one function and three accessors. It hands you the official client for each resource — a real @libsql/client, ioredis, or @aws-sdk/client-s3 instance — already pointed at the right endpoint and scoped to your project. There is no Flect-specific query wrapper to learn.

npm install @getflect/sdk

createEnv(options?)

import { createEnv } from '@getflect/sdk'

const env = createEnv()

createEnv() is I/O-free — it does no network calls until you resolve a binding. By default it reads:

• FLECT_BROKER_URL — the broker to resolve bindings through
• FLECT_TOKEN — the runtime token (both injected automatically on deploy)

When no broker URL is configured, it switches to local mode and resolves from flect.local.json (see Local development).

Options

Option	Description
brokerUrl	Broker base URL. Default: FLECT_BROKER_URL, else http://localhost:8787.
authToken	Runtime token. Default: FLECT_TOKEN.
local	true to force local mode (flect.local.json / FLECT_LOCAL_CONFIG), or a config object.
retry	Retry policy for broker resolution calls.
fetch	Custom fetch implementation (tests, proxies).

Each accessor returns a Promise — resolution happens on first use, and clients are cached until their lease expires.

---

Database — env.db(binding)

Returns an official @libsql/client instance. Install it as a peer dependency: npm install @libsql/client.

const db = await env.db('DB')

// libsql API — parameterized queries, batches, transactions
const { rows } = await db.execute('SELECT id, title FROM notes ORDER BY created_at DESC')

await db.execute({
  sql: 'INSERT INTO notes (id, title) VALUES (?, ?)',
  args: [crypto.randomUUID(), 'Hello'],
})

await db.batch([
  { sql: 'INSERT INTO posts (id, title) VALUES (?, ?)', args: [id, title] },
  { sql: 'UPDATE users SET post_count = post_count + 1 WHERE id = ?', args: [userId] },
])

Isolation is transparent: Flect stamps the resource’s sqld namespace on every request, so a binding only ever sees its own data. Because it’s the standard client, it drops straight into drizzle-orm/libsql:

import { drizzle } from 'drizzle-orm/libsql'
const orm = drizzle(await env.db('DB'))

---

KV — env.kv(binding)

Returns an official ioredis instance. Install it as a peer dependency: npm install ioredis.

const cache = await env.kv('CACHE')

await cache.set('session:abc', 'active', 'EX', 3600)   // 1h TTL
const val = await cache.get('session:abc')
await cache.del('session:abc')

await cache.incr('counter')
const many = await cache.mget('a', 'b', 'c')

Every key is transparently prefixed with the resource’s namespace, so stores never collide. The full ioredis command set is available.

---

Object storage — env.store(binding)

Returns an official @aws-sdk/client-s3 S3Client, pre-configured with the resource’s endpoint, region, bucket, and credentials. Install the peers: npm install @aws-sdk/client-s3.

import { PutObjectCommand, GetObjectCommand } from '@aws-sdk/client-s3'

const s3 = await env.store('FILES')
const bucket = process.env.FILES_BUCKET   // the resolved bucket name

await s3.send(new PutObjectCommand({
  Bucket: bucket,
  Key: 'avatars/alice.png',
  Body: buffer,
  ContentType: 'image/png',
}))

const obj = await s3.send(new GetObjectCommand({ Bucket: bucket, Key: 'avatars/alice.png' }))

For presigned URLs, use @aws-sdk/s3-request-presigner with the same client — it’s a standard S3Client.

---

Typing the client

Each accessor takes an optional type parameter if you want to annotate the returned client:

import type { Client } from '@libsql/client'
const db = await env.db<Client>('DB')

---

Why it works this way

Your code depends only on well-known, well-documented libraries — Flect owns none of your data-access API. The platform’s job is resolution and isolation: turning env.db('DB') into a scoped, short-lived, correctly-namespaced client, without a URL or secret ever appearing in your source. See Platform architecture for how resolution works.