fauxqs 1.12.0 → 2.0.0

package/README.md

@@ -1,1462 +1,1532 @@
-<img src="logo-readme.jpg" alt="fauxqs" width="360" />
-
-[![npm version](https://img.shields.io/npm/v/fauxqs.svg)](https://www.npmjs.com/package/fauxqs)
-
-# fauxqs
-
-Local SNS/SQS/S3 emulator for development and testing. Point your `@aws-sdk/client-sqs`, `@aws-sdk/client-sns`, and `@aws-sdk/client-s3` clients at fauxqs instead of real AWS.
-
-All state is in-memory. No persistence, no external storage dependencies.
-
-## Table of Contents
-
-- [Installation](#installation)
-- [Usage](#usage)
-  - [Running the server](#running-the-server)
-  - [Running in the background](#running-in-the-background)
-  - [Running with Docker](#running-with-docker)
-  - [Running in Docker Compose](#running-in-docker-compose)
-    - [Container-to-container S3 virtual-hosted-style](#container-to-container-s3-virtual-hosted-style)
-  - [Configuring AWS SDK clients](#configuring-aws-sdk-clients)
-  - [Programmatic usage](#programmatic-usage)
-    - [Relaxed rules](#relaxed-rules)
-    - [Programmatic state setup](#programmatic-state-setup)
-    - [Sending messages programmatically](#sending-messages-programmatically)
-    - [Init config file](#init-config-file)
-    - [Init config schema reference](#init-config-schema-reference)
-    - [Message spy](#message-spy)
-    - [Queue inspection](#queue-inspection)
-  - [Configurable queue URL host](#configurable-queue-url-host)
-  - [Region](#region)
-- [Supported API Actions](#supported-api-actions)
-  - [SQS](#sqs)
-  - [SNS](#sns)
-  - [S3](#s3)
-  - [STS](#sts)
-- [SQS Features](#sqs-features)
-- [SNS Features](#sns-features)
-- [S3 Features](#s3-features)
-  - [S3 URL styles](#s3-url-styles)
-  - [Using with AWS CLI](#using-with-aws-cli)
-- [Testing Strategies](#testing-strategies)
-  - [Library mode for tests](#library-mode-for-tests)
-  - [Docker mode for local development](#docker-mode-for-local-development)
-  - [Recommended combination](#recommended-combination)
-- [Conventions](#conventions)
-- [Limitations](#limitations)
-- [Examples](#examples)
-- [Migrating from LocalStack](#migrating-from-localstack)
-  - [SNS/SQS only](#snssqs-only)
-  - [SNS/SQS/S3](#snssqss3)
-  - [Going hybrid (recommended)](#going-hybrid-recommended)
-- [Benchmarks](#benchmarks)
-- [License](#license)
-
-## Installation
-
-**Docker** (recommended for standalone usage) — [Docker Hub](https://hub.docker.com/r/kibertoad/fauxqs):
-
-```bash
-docker run -p 4566:4566 kibertoad/fauxqs
-```
-
-**npm** (for embedded library usage or CLI):
-
-```bash
-npm install fauxqs
-```
-
-## Usage
-
-### Running the server
-
-```bash
-npx fauxqs
-```
-
-The server starts on port `4566` and handles SQS, SNS, and S3 on a single endpoint.
-
-#### Environment variables
-
-| Variable | Description | Default |
-|----------|-------------|---------|
-| `FAUXQS_PORT` | Port to listen on | `4566` |
-| `FAUXQS_HOST` | Host for queue URLs (`sqs.<region>.<host>` format) | `localhost` |
-| `FAUXQS_DEFAULT_REGION` | Fallback region for ARNs and URLs | `us-east-1` |
-| `FAUXQS_LOGGER` | Enable request logging (`true`/`false`) | `true` |
-| `FAUXQS_INIT` | Path to a JSON init config file (see [Init config file](#init-config-file)) | (none) |
-| `FAUXQS_DNS_NAME` | Domain that dnsmasq resolves (including all subdomains) to the container IP. Only needed when the container hostname doesn't match the docker-compose service name — e.g., when using `container_name` or running with plain `docker run`. In docker-compose the hostname is set to the service name automatically, so this is rarely needed. (Docker only) | container hostname |
-| `FAUXQS_DNS_UPSTREAM` | Where dnsmasq forwards non-fauxqs DNS queries (e.g., `registry.npmjs.org`). Change this if you're in a corporate network with an internal DNS server, or if you prefer a different public resolver like `1.1.1.1`. (Docker only) | `8.8.8.8` |
-
-```bash
-FAUXQS_PORT=3000 FAUXQS_INIT=init.json npx fauxqs
-```
-
-A health check is available at `GET /health`.
-
-### Running in the background
-
-To keep fauxqs running while you work on your app or run tests repeatedly, start it as a background process:
-
-```bash
-npx fauxqs &
-```
-
-Or in a separate terminal:
-
-```bash
-npx fauxqs
-```
-
-All state accumulates in memory across requests, so queues, topics, and objects persist until the server is stopped.
-
-To stop the server:
-
-```bash
-# If backgrounded in the same shell
-kill %1
-
-# Cross-platform, by port
-npx cross-port-killer 4566
-```
-
-### Running with Docker
-
-The official Docker image is available on Docker Hub:
-
-```bash
-docker run -p 4566:4566 kibertoad/fauxqs
-```
-
-With an init config file:
-
-```bash
-docker run -p 4566:4566 \
-  -v ./init.json:/app/init.json \
-  -e FAUXQS_INIT=/app/init.json \
-  kibertoad/fauxqs
-```
-
-### Running in Docker Compose
-
-Use the `kibertoad/fauxqs` image and mount a JSON init config to pre-create resources on startup:
-
-```json
-// scripts/fauxqs/init.json
-{
-  "queues": [
-    {
-      "name": "my-queue.fifo",
-      "attributes": { "FifoQueue": "true", "ContentBasedDeduplication": "true" }
-    },
-    { "name": "my-dlq" }
-  ],
-  "topics": [{ "name": "my-events" }],
-  "subscriptions": [{ "topic": "my-events", "queue": "my-dlq" }],
-  "buckets": ["my-uploads"]
-}
-```
-
-```yaml
-# docker-compose.yml
-services:
-  fauxqs:
-    image: kibertoad/fauxqs:latest
-    ports:
-      - "4566:4566"
-    environment:
-      - FAUXQS_INIT=/app/init.json
-    volumes:
-      - ./scripts/fauxqs/init.json:/app/init.json
-
-  app:
-    # ...
-    depends_on:
-      fauxqs:
-        condition: service_healthy
-```
-
-The image has a built-in `HEALTHCHECK`, so `service_healthy` works without extra configuration in your compose file. Other containers reference fauxqs using the Docker service name (`http://fauxqs:4566`). The init config file creates all queues, topics, subscriptions, and buckets before the healthcheck passes, so dependent services start only after resources are ready.
-
-#### Container-to-container S3 virtual-hosted-style
-
-The Docker image includes a built-in DNS server ([dnsmasq](https://thekelleys.org.uk/dnsmasq/doc.html)) that resolves the container hostname and all its subdomains (e.g., `fauxqs`, `s3.fauxqs`, `my-bucket.s3.fauxqs`) to the container's own IP. This enables virtual-hosted-style S3 from other containers without `forcePathStyle`.
-
-To use it, assign fauxqs a static IP and point other containers' DNS to it:
-
-```yaml
-# docker-compose.yml
-services:
-  fauxqs:
-    image: kibertoad/fauxqs:latest
-    networks:
-      default:
-        ipv4_address: 10.0.0.2
-    ports:
-      - "4566:4566"
-    environment:
-      - FAUXQS_INIT=/app/init.json
-      - FAUXQS_HOST=fauxqs
-    volumes:
-      - ./scripts/fauxqs/init.json:/app/init.json
-
-  app:
-    dns: 10.0.0.2
-    depends_on:
-      fauxqs:
-        condition: service_healthy
-    environment:
-      - AWS_ENDPOINT=http://s3.fauxqs:4566
-
-networks:
-  default:
-    ipam:
-      config:
-        - subnet: 10.0.0.0/24
-```
-
-From the `app` container, `my-bucket.s3.fauxqs` resolves to `10.0.0.2` (the fauxqs container), so virtual-hosted-style S3 works:
-
-```typescript
-const s3 = new S3Client({
-  endpoint: "http://s3.fauxqs:4566",
-  region: "us-east-1",
-  credentials: { accessKeyId: "test", secretAccessKey: "test" },
-  // No forcePathStyle needed!
-});
-```
-
-The DNS server is configured automatically using the container hostname (which docker-compose sets to the service name), so in most setups no extra configuration is needed. See the [environment variables table](#environment-variables) for `FAUXQS_DNS_NAME` and `FAUXQS_DNS_UPSTREAM` if you need to override the defaults.
-
-### Configuring AWS SDK clients
-
-Point your SDK clients at the local server:
-
-```typescript
-import { SQSClient } from "@aws-sdk/client-sqs";
-import { SNSClient } from "@aws-sdk/client-sns";
-import { S3Client } from "@aws-sdk/client-s3";
-
-const sqsClient = new SQSClient({
-  endpoint: "http://localhost:4566",
-  region: "us-east-1",
-  credentials: { accessKeyId: "test", secretAccessKey: "test" },
-});
-
-const snsClient = new SNSClient({
-  endpoint: "http://localhost:4566",
-  region: "us-east-1",
-  credentials: { accessKeyId: "test", secretAccessKey: "test" },
-});
-
-// Using fauxqs.dev wildcard DNS — no helpers or forcePathStyle needed
-const s3Client = new S3Client({
-  endpoint: "http://s3.localhost.fauxqs.dev:4566",
-  region: "us-east-1",
-  credentials: { accessKeyId: "test", secretAccessKey: "test" },
-});
-```
-
-Any credentials are accepted and never validated.
-
-> **Note:** The `fauxqs.dev` wildcard DNS (`*.localhost.fauxqs.dev` → `127.0.0.1`) replicates the approach [pioneered by LocalStack](https://hashnode.localstack.cloud/efficient-localstack-s3-endpoint-configuration) with `localhost.localstack.cloud`. A public DNS entry resolves all subdomains to localhost, so virtual-hosted-style S3 requests work without `/etc/hosts` changes, custom request handlers, or `forcePathStyle`. See [S3 URL styles](#s3-url-styles) for alternative approaches.
-
-### Programmatic usage
-
-You can also embed fauxqs directly in your test suite:
-
-```typescript
-import { startFauxqs } from "fauxqs";
-
-const server = await startFauxqs({ port: 4566, logger: false });
-
-console.log(server.address); // "http://127.0.0.1:4566"
-console.log(server.port);    // 4566
-
-// point your SDK clients at server.address
-
-// clean up when done
-await server.stop();
-```
-
-Pass `port: 0` to let the OS assign a random available port (useful in tests).
-
-#### Relaxed rules
-
-By default, fauxqs enforces AWS-strict validations. You can selectively relax some of these for convenience during development:
-
-```typescript
-const server = await startFauxqs({
-  port: 0,
-  relaxedRules: {
-    disableMinCopySourceSize: true,
-  },
-});
-```
-
-| Rule | Default | Description |
-|------|---------|-------------|
-| `disableMinCopySourceSize` | `false` | AWS requires the source object to be [larger than 5 MiB](https://docs.aws.amazon.com/AmazonS3/latest/API/API_UploadPartCopy.html) for byte-range `UploadPartCopy`. Set to `true` to allow byte-range copies from smaller sources. |
-
-#### Programmatic state setup
-
-The server object exposes methods for pre-creating resources without going through the SDK:
-
-```typescript
-const server = await startFauxqs({ port: 0, logger: false });
-
-// Create individual resources — each returns metadata about the created resource
-const { queueUrl, queueArn, queueName } = server.createQueue("my-queue");
-const { queueUrl: dlqUrl } = server.createQueue("my-dlq", {
-  attributes: { VisibilityTimeout: "60" },
-  tags: { env: "test" },
-});
-const { topicArn } = server.createTopic("my-topic");
-server.subscribe({ topic: "my-topic", queue: "my-queue" });
-const { bucketName } = server.createBucket("my-bucket");
-
-// Create resources in a specific region
-server.createQueue("eu-queue", { region: "eu-west-1" });
-server.createTopic("eu-topic", { region: "eu-west-1" });
-server.subscribe({ topic: "eu-topic", queue: "eu-queue", region: "eu-west-1" });
-
-// Or create everything at once
-server.setup({
-  queues: [
-    { name: "orders" },
-    { name: "notifications", attributes: { DelaySeconds: "5" } },
-    { name: "eu-orders", region: "eu-west-1" },
-  ],
-  topics: [{ name: "events" }],
-  subscriptions: [
-    { topic: "events", queue: "orders" },
-    { topic: "events", queue: "notifications" },
-  ],
-  buckets: ["uploads", "exports"],
-});
-
-// Delete individual resources (uses defaultRegion; pass { region } to override)
-server.deleteQueue("my-queue");                          // no-op if queue doesn't exist
-server.deleteQueue("eu-queue", { region: "eu-west-1" }); // delete in specific region
-server.deleteTopic("my-topic");                          // also removes associated subscriptions
-server.deleteTopic("eu-topic", { region: "eu-west-1" });
-server.emptyBucket("my-bucket");                         // removes all objects, keeps the bucket
-
-// Clear all messages and S3 objects between tests (keeps queues, topics, subscriptions, buckets)
-server.reset();
-
-// Or nuke everything — removes queues, topics, subscriptions, and buckets too
-server.purgeAll();
-```
-
-#### Sending messages programmatically
-
-Send SQS messages and publish to SNS topics without instantiating SDK clients:
-
-```typescript
-// SQS: enqueue a message directly
-const { messageId, md5OfBody } = server.sendMessage("my-queue", "hello world");
-
-// With message attributes
-server.sendMessage("my-queue", JSON.stringify({ orderId: "123" }), {
-  messageAttributes: {
-    eventType: { DataType: "String", StringValue: "order.created" },
-  },
-});
-
-// With delay
-server.sendMessage("my-queue", "delayed message", { delaySeconds: 10 });
-
-// FIFO queue — returns sequenceNumber
-const { sequenceNumber } = server.sendMessage("my-queue.fifo", "fifo message", {
-  messageGroupId: "group-1",
-  messageDeduplicationId: "dedup-1",
-});
-
-// SNS: publish to a topic (fans out to all SQS subscriptions)
-const { messageId: snsMessageId } = server.publish("my-topic", "event payload");
-
-// With subject and message attributes
-server.publish("my-topic", JSON.stringify({ orderId: "456" }), {
-  subject: "Order Update",
-  messageAttributes: {
-    eventType: { DataType: "String", StringValue: "order.updated" },
-  },
-});
-
-// FIFO topic
-server.publish("my-topic.fifo", "fifo event", {
-  messageGroupId: "group-1",
-  messageDeduplicationId: "dedup-1",
-});
-```
-
-`sendMessage` validates the message body (invalid characters, size limits), applies queue-level `DelaySeconds` defaults, handles FIFO deduplication, and emits spy events automatically. Returns `{ messageId, md5OfBody, md5OfMessageAttributes?, sequenceNumber? }`.
-
-`publish` validates message size, evaluates filter policies on each subscription, supports raw message delivery, and emits spy events. Returns `{ messageId }`.
-
-Both methods throw if the target queue or topic doesn't exist.
-
-#### Init config file
-
-Create a JSON file to pre-create resources on startup. The file is validated on load — malformed configs produce a clear error instead of silent failures.
-
-```json
-{
-  "queues": [
-    { "name": "orders" },
-    { "name": "orders-dlq" },
-    { "name": "orders.fifo", "attributes": { "FifoQueue": "true", "ContentBasedDeduplication": "true" } }
-  ],
-  "topics": [
-    { "name": "events" }
-  ],
-  "subscriptions": [
-    { "topic": "events", "queue": "orders" }
-  ],
-  "buckets": ["uploads", "exports"]
-}
-```
-
-Pass it via the `FAUXQS_INIT` environment variable or the `init` option:
-
-```bash
-FAUXQS_INIT=init.json npx fauxqs
-```
-
-```typescript
-const server = await startFauxqs({ init: "init.json" });
-// or inline:
-const server = await startFauxqs({
-  init: { queues: [{ name: "my-queue" }], buckets: ["my-bucket"] },
-});
-```
-
-#### Init config schema reference
-
-All top-level fields are optional. Resources are created in dependency order: queues, topics, subscriptions, buckets.
-
-##### `queues`
-
-Array of queue objects.
-
-| Field | Type | Required | Description |
-|-------|------|----------|-------------|
-| `name` | `string` | Yes | Queue name. Use `.fifo` suffix for FIFO queues. |
-| `region` | `string` | No | Override the default region for this queue. The queue's ARN and URL will use this region. |
-| `attributes` | `Record<string, string>` | No | Queue attributes (see table below). |
-| `tags` | `Record<string, string>` | No | Key-value tags for the queue. |
-
-Supported queue attributes:
-
-| Attribute | Default | Range / Values |
-|-----------|---------|----------------|
-| `VisibilityTimeout` | `"30"` | `0` – `43200` (seconds) |
-| `DelaySeconds` | `"0"` | `0` – `900` (seconds) |
-| `MaximumMessageSize` | `"1048576"` | `1024` – `1048576` (bytes) |
-| `MessageRetentionPeriod` | `"345600"` | `60` – `1209600` (seconds) |
-| `ReceiveMessageWaitTimeSeconds` | `"0"` | `0` – `20` (seconds) |
-| `RedrivePolicy` | — | JSON string: `{"deadLetterTargetArn": "arn:...", "maxReceiveCount": "5"}` |
-| `Policy` | — | Queue policy JSON string (stored, not enforced) |
-| `KmsMasterKeyId` | — | KMS key ID (stored, no actual encryption) |
-| `KmsDataKeyReusePeriodSeconds` | — | KMS data key reuse period (stored, no actual encryption) |
-| `FifoQueue` | — | `"true"` for FIFO queues (queue name must end with `.fifo`) |
-| `ContentBasedDeduplication` | — | `"true"` or `"false"` (FIFO queues only) |
-
-Example:
-
-```json
-{
-  "queues": [
-    {
-      "name": "orders",
-      "attributes": { "VisibilityTimeout": "60", "DelaySeconds": "5" },
-      "tags": { "env": "staging", "team": "platform" }
-    },
-    {
-      "name": "orders-dlq"
-    },
-    {
-      "name": "orders.fifo",
-      "attributes": {
-        "FifoQueue": "true",
-        "ContentBasedDeduplication": "true"
-      }
-    },
-    {
-      "name": "retry-queue",
-      "attributes": {
-        "RedrivePolicy": "{\"deadLetterTargetArn\":\"arn:aws:sqs:us-east-1:000000000000:orders-dlq\",\"maxReceiveCount\":\"3\"}"
-      }
-    },
-    {
-      "name": "eu-orders",
-      "region": "eu-west-1"
-    }
-  ]
-}
-```
-
-##### `topics`
-
-Array of topic objects.
-
-| Field | Type | Required | Description |
-|-------|------|----------|-------------|
-| `name` | `string` | Yes | Topic name. Use `.fifo` suffix for FIFO topics. |
-| `region` | `string` | No | Override the default region for this topic. The topic's ARN will use this region. |
-| `attributes` | `Record<string, string>` | No | Topic attributes (e.g., `DisplayName`). |
-| `tags` | `Record<string, string>` | No | Key-value tags for the topic. |
-
-Example:
-
-```json
-{
-  "topics": [
-    {
-      "name": "events",
-      "attributes": { "DisplayName": "Application Events" },
-      "tags": { "env": "staging" }
-    },
-    {
-      "name": "events.fifo",
-      "attributes": { "FifoQueue": "true", "ContentBasedDeduplication": "true" }
-    }
-  ]
-}
-```
-
-##### `subscriptions`
-
-Array of subscription objects. Referenced topics and queues must be defined in the same config (or already exist on the server).
-
-| Field | Type | Required | Description |
-|-------|------|----------|-------------|
-| `topic` | `string` | Yes | Topic name (not ARN) to subscribe to. |
-| `queue` | `string` | Yes | Queue name (not ARN) to deliver messages to. |
-| `region` | `string` | No | Override the default region. The topic and queue ARNs will be resolved in this region. |
-| `attributes` | `Record<string, string>` | No | Subscription attributes (see table below). |
-
-Supported subscription attributes:
-
-| Attribute | Values | Description |
-|-----------|--------|-------------|
-| `RawMessageDelivery` | `"true"` / `"false"` | Deliver the raw message body instead of the SNS envelope JSON. |
-| `FilterPolicy` | JSON string | SNS filter policy for message filtering (e.g., `"{\"color\": [\"blue\"]}"`) |
-| `FilterPolicyScope` | `"MessageAttributes"` / `"MessageBody"` | Whether the filter policy applies to message attributes or body. Defaults to `MessageAttributes`. |
-| `RedrivePolicy` | JSON string | Subscription-level dead-letter queue config. |
-| `DeliveryPolicy` | JSON string | Delivery retry policy (stored, not enforced). |
-| `SubscriptionRoleArn` | ARN string | IAM role ARN for delivery (stored, not enforced). |
-
-Example:
-
-```json
-{
-  "subscriptions": [
-    {
-      "topic": "events",
-      "queue": "orders",
-      "attributes": {
-        "RawMessageDelivery": "true",
-        "FilterPolicy": "{\"eventType\": [\"order.created\", \"order.updated\"]}"
-      }
-    },
-    {
-      "topic": "events",
-      "queue": "notifications"
-    }
-  ]
-}
-```
-
-##### `buckets`
-
-Array of bucket name strings.
-
-```json
-{
-  "buckets": ["uploads", "exports", "temp"]
-}
-```
-
-#### Message spy
-
-`MessageSpyReader` lets you await specific events flowing through SQS, SNS, and S3 in your tests — without polling queues yourself. Inspired by `HandlerSpy` from `message-queue-toolkit`.
-
-Enable it with the `messageSpies` option:
-
-```typescript
-const server = await startFauxqs({ port: 0, logger: false, messageSpies: true });
-```
-
-The spy tracks events across all three services using a discriminated union on `service`:
-
-**SQS events** (`service: 'sqs'`):
-- **`published`** — message was enqueued (via SendMessage, SendMessageBatch, or SNS fan-out)
-- **`consumed`** — message was deleted (via DeleteMessage / DeleteMessageBatch)
-- **`dlq`** — message exceeded `maxReceiveCount` and was moved to a dead-letter queue
-
-**SNS events** (`service: 'sns'`):
-- **`published`** — message was published to a topic (before fan-out to SQS subscriptions)
-
-**S3 events** (`service: 's3'`):
-- **`uploaded`** — object was put (PutObject or CompleteMultipartUpload)
-- **`downloaded`** — object was retrieved (GetObject)
-- **`deleted`** — object was deleted (DeleteObject, only when key existed)
-- **`copied`** — object was copied (CopyObject; also emits `uploaded` for the destination)
-
-##### Awaiting messages
-
-```typescript
-// Wait for a specific SQS message (resolves immediately if already in buffer)
-const msg = await server.spy.waitForMessage(
-  (m) => m.service === "sqs" && m.body === "order.created" && m.queueName === "orders",
-  "published",
-);
-
-// Wait by SQS message ID
-const msg = await server.spy.waitForMessageWithId(messageId, "consumed");
-
-// Partial object match (deep-equal on specified fields)
-const msg = await server.spy.waitForMessage({ service: "sqs", queueName: "orders", status: "published" });
-
-// Wait for an SNS publish event
-const msg = await server.spy.waitForMessage({ service: "sns", topicName: "my-topic", status: "published" });
-
-// Wait for an S3 upload event
-const msg = await server.spy.waitForMessage({ service: "s3", bucket: "my-bucket", key: "file.txt", status: "uploaded" });
-```
-
-`waitForMessage` checks the buffer first (retroactive resolution). If no match is found, it returns a Promise that resolves when a matching message arrives.
-
-##### Timeout
-
-All `waitForMessage` and `waitForMessageWithId` calls accept an optional `timeout` parameter (ms) as the third argument. If no matching message arrives in time, the promise rejects with a timeout error — preventing tests from hanging indefinitely:
-
-```typescript
-// Reject after 2 seconds if no match
-const msg = await server.spy.waitForMessage(
-  { service: "sqs", queueName: "orders" },
-  "published",
-  2000,
-);
-
-// Also works with waitForMessageWithId
-const msg = await server.spy.waitForMessageWithId(messageId, "consumed", 5000);
-```
-
-##### Waiting for multiple messages
-
-`waitForMessages` collects `count` matching messages before resolving. It checks the buffer first, then awaits future arrivals:
-
-```typescript
-// Wait for 3 messages on the orders queue
-const msgs = await server.spy.waitForMessages(
-  { service: "sqs", queueName: "orders" },
-  { count: 3, status: "published", timeout: 5000 },
-);
-// msgs.length === 3
-```
-
-If the timeout expires before enough messages arrive, the promise rejects with a message showing how many were collected (e.g., `"collected 1/3"`).
-
-##### Negative assertions
-
-`expectNoMessage` asserts that no matching message appears within a time window. Useful for verifying that filter policies dropped a message or that a side effect did not occur:
-
-```typescript
-// Assert no message was delivered to the wrong queue (waits 200ms by default)
-await server.spy.expectNoMessage({ service: "sqs", queueName: "wrong-queue" });
-
-// Custom window and status filter
-await server.spy.expectNoMessage(
-  { service: "sqs", queueName: "orders" },
-  { status: "dlq", within: 500 },
-);
-```
-
-If a matching message is already in the buffer, `expectNoMessage` rejects immediately. If one arrives during the wait, it rejects with `"matching message arrived during wait"`.
-
-##### Synchronous check
-
-```typescript
-const msg = server.spy.checkForMessage(
-  (m) => m.service === "sqs" && m.queueName === "my-queue",
-  "published",
-);
-// returns SpyMessage | undefined
-```
-
-##### Buffer management
-
-```typescript
-// Get all tracked messages (oldest to newest)
-const all = server.spy.getAllMessages();
-
-// Clear buffer and reject pending waiters
-server.spy.clear();
-```
-
-The buffer defaults to 100 messages (FIFO eviction). Configure with:
-
-```typescript
-const server = await startFauxqs({
-  messageSpies: { bufferSize: 500 },
-});
-```
-
-##### Types
-
-`server.spy` returns a `MessageSpyReader` — a read-only interface that exposes query and await methods but not internal mutation (e.g. recording new events):
-
-```typescript
-interface MessageSpyReader {
-  waitForMessage(filter: MessageSpyFilter, status?: string, timeout?: number): Promise<SpyMessage>;
-  waitForMessageWithId(messageId: string, status?: string, timeout?: number): Promise<SpyMessage>;
-  waitForMessages(filter: MessageSpyFilter, options: WaitForMessagesOptions): Promise<SpyMessage[]>;
-  expectNoMessage(filter: MessageSpyFilter, options?: ExpectNoMessageOptions): Promise<void>;
-  checkForMessage(filter: MessageSpyFilter, status?: string): SpyMessage | undefined;
-  getAllMessages(): SpyMessage[];
-  clear(): void;
-}
-
-interface WaitForMessagesOptions {
-  count: number;
-  status?: string;
-  timeout?: number;
-}
-
-interface ExpectNoMessageOptions {
-  status?: string;
-  within?: number; // ms, defaults to 200
-}
-```
-
-`SpyMessage` is a discriminated union:
-
-```typescript
-interface SqsSpyMessage {
-  service: "sqs";
-  queueName: string;
-  messageId: string;
-  body: string;
-  messageAttributes: Record<string, MessageAttributeValue>;
-  status: "published" | "consumed" | "dlq";
-  timestamp: number;
-}
-
-interface SnsSpyMessage {
-  service: "sns";
-  topicArn: string;
-  topicName: string;
-  messageId: string;
-  body: string;
-  messageAttributes: Record<string, MessageAttributeValue>;
-  status: "published";
-  timestamp: number;
-}
-
-interface S3SpyEvent {
-  service: "s3";
-  bucket: string;
-  key: string;
-  status: "uploaded" | "downloaded" | "deleted" | "copied";
-  timestamp: number;
-}
-
-type SpyMessage = SqsSpyMessage | SnsSpyMessage | S3SpyEvent;
-```
-
-##### Spy disabled by default
-
-Accessing `server.spy` when `messageSpies` is not set throws an error. There is no overhead on the message flow when spies are disabled.
-
-#### Queue inspection
-
-Non-destructive inspection of SQS queue state — see all messages (ready, in-flight, and delayed) without consuming them or affecting visibility timeouts.
-
-##### Programmatic API
-
-```typescript
-const result = server.inspectQueue("my-queue");
-// result is undefined if queue doesn't exist
-if (result) {
-  console.log(result.name);           // "my-queue"
-  console.log(result.url);            // "http://sqs.us-east-1.localhost:4566/000000000000/my-queue"
-  console.log(result.arn);            // "arn:aws:sqs:us-east-1:000000000000:my-queue"
-  console.log(result.attributes);     // { VisibilityTimeout: "30", ... }
-  console.log(result.messages.ready);    // messages available for receive
-  console.log(result.messages.delayed);  // messages waiting for delay to expire
-  console.log(result.messages.inflight); // received but not yet deleted
-  // Each inflight entry includes: { message, receiptHandle, visibilityDeadline }
-}
-```
-
-##### HTTP endpoints
-
-```bash
-# List all queues with summary counts
-curl http://localhost:4566/_fauxqs/queues
-# [{ "name": "my-queue", "approximateMessageCount": 5, "approximateInflightCount": 2, "approximateDelayedCount": 0, ... }]
-
-# Inspect a specific queue (full state)
-curl http://localhost:4566/_fauxqs/queues/my-queue
-# { "name": "my-queue", "messages": { "ready": [...], "delayed": [...], "inflight": [...] }, ... }
-```
-
-Returns 404 for non-existent queues. Inspection never modifies queue state — messages remain exactly where they are.
-
-### Configurable queue URL host
-
-Queue URLs use the AWS-style `sqs.<region>.<host>` format. The `host` defaults to `localhost`, producing URLs like `http://sqs.us-east-1.localhost:4566/000000000000/myQueue`.
-
-To override the host (e.g., for a custom domain):
-
-```typescript
-import { startFauxqs } from "fauxqs";
-
-const server = await startFauxqs({ port: 4566, host: "myhost.local" });
-// Queue URLs: http://sqs.us-east-1.myhost.local:4566/000000000000/myQueue
-```
-
-This also works with `buildApp`:
-
-```typescript
-import { buildApp } from "fauxqs";
-
-const app = buildApp({ host: "myhost.local" });
-```
-
-The configured host ensures queue URLs are consistent across all creation paths (init config, programmatic API, and SDK requests), regardless of the request's `Host` header.
-
-### Region
-
-Region is part of an entity's identity — a queue named `my-queue` in `us-east-1` is a completely different entity from `my-queue` in `eu-west-1`, just like in real AWS.
-
-The region used in ARNs and queue URLs is automatically detected from the SDK client's `Authorization` header (AWS SigV4 credential scope). If your SDK client is configured with `region: "eu-west-1"`, all entities created or looked up through that client will use `eu-west-1` in their ARNs and URLs.
-
-```typescript
-const sqsEU = new SQSClient({ region: "eu-west-1", endpoint: "http://localhost:4566", ... });
-const sqsUS = new SQSClient({ region: "us-east-1", endpoint: "http://localhost:4566", ... });
-
-// These are two independent queues with different ARNs
-await sqsEU.send(new CreateQueueCommand({ QueueName: "orders" }));
-await sqsUS.send(new CreateQueueCommand({ QueueName: "orders" }));
-```
-
-If the region cannot be resolved from request headers (e.g., requests without AWS SigV4 signing), the `defaultRegion` option is used as a fallback (defaults to `"us-east-1"`):
-
-```typescript
-const server = await startFauxqs({ defaultRegion: "eu-west-1" });
-```
-
-Resources created via init config or programmatic API use the `defaultRegion` unless overridden with an explicit `region` field:
-
-```json
-{
-  "queues": [
-    { "name": "us-queue" },
-    { "name": "eu-queue", "region": "eu-west-1" }
-  ]
-}
-```
-
-## Supported API Actions
-
-### SQS
-
-| Action | Supported |
-|--------|-----------|
-| CreateQueue | Yes |
-| DeleteQueue | Yes |
-| GetQueueUrl | Yes |
-| ListQueues | Yes |
-| GetQueueAttributes | Yes |
-| SetQueueAttributes | Yes |
-| PurgeQueue | Yes |
-| SendMessage | Yes |
-| SendMessageBatch | Yes |
-| ReceiveMessage | Yes |
-| DeleteMessage | Yes |
-| DeleteMessageBatch | Yes |
-| ChangeMessageVisibility | Yes |
-| ChangeMessageVisibilityBatch | Yes |
-| TagQueue | Yes |
-| UntagQueue | Yes |
-| ListQueueTags | Yes |
-| AddPermission | No |
-| RemovePermission | No |
-| ListDeadLetterSourceQueues | No |
-| StartMessageMoveTask | No |
-| CancelMessageMoveTask | No |
-| ListMessageMoveTasks | No |
-
-### SNS
-
-| Action | Supported |
-|--------|-----------|
-| CreateTopic | Yes |
-| DeleteTopic | Yes |
-| ListTopics | Yes |
-| GetTopicAttributes | Yes |
-| SetTopicAttributes | Yes |
-| Subscribe | Yes |
-| Unsubscribe | Yes |
-| ConfirmSubscription | Yes |
-| ListSubscriptions | Yes |
-| ListSubscriptionsByTopic | Yes |
-| GetSubscriptionAttributes | Yes |
-| SetSubscriptionAttributes | Yes |
-| Publish | Yes |
-| PublishBatch | Yes |
-| TagResource | Yes |
-| UntagResource | Yes |
-| ListTagsForResource | Yes |
-| AddPermission | No |
-| RemovePermission | No |
-| GetDataProtectionPolicy | No |
-| PutDataProtectionPolicy | No |
-
-Platform application, SMS, and phone number actions are not supported.
-
-### S3
-
-| Action | Supported |
-|--------|-----------|
-| CreateBucket | Yes |
-| HeadBucket | Yes |
-| DeleteBucket | Yes |
-| ListBuckets | Yes |
-| ListObjects | Yes |
-| ListObjectsV2 | Yes |
-| PutObject | Yes |
-| GetObject | Yes |
-| HeadObject | Yes |
-| DeleteObject | Yes |
-| DeleteObjects | Yes |
-| CopyObject | Yes |
-| CreateMultipartUpload | Yes |
-| UploadPart | Yes |
-| UploadPartCopy | Yes |
-| CompleteMultipartUpload | Yes |
-| AbortMultipartUpload | Yes |
-| GetObjectAttributes | Yes |
-| GetBucketLocation | No |
-| ListObjectVersions | No |
-| SelectObjectContent | No |
-| RestoreObject | No |
-| RenameObject | No |
-| ListMultipartUploads | No |
-| ListParts | No |
-
-Bucket configuration (CORS, lifecycle, encryption, replication, logging, website, notifications, policy), ACLs, versioning, tagging, object lock, and public access block actions are not supported.
-
-### STS
-
-| Action | Supported |
-|--------|-----------|
-| GetCallerIdentity | Yes |
-| AssumeRole | No |
-| GetSessionToken | No |
-| GetFederationToken | No |
-
-Returns a mock identity with account `000000000000` and ARN `arn:aws:iam::000000000000:root`. This allows tools like Terraform and the AWS CLI that call `sts:GetCallerIdentity` on startup to work without errors. Other STS actions are not supported.
-
-## SQS Features
-
-- **Message attributes** with MD5 checksums matching the AWS algorithm
-- **Visibility timeout** — messages become invisible after receive and reappear after timeout
-- **Delay queues** — per-queue default delay and per-message delay overrides
-- **Long polling** — `WaitTimeSeconds` on ReceiveMessage blocks until messages arrive or timeout
-- **Dead letter queues** — messages exceeding `maxReceiveCount` are moved to the configured DLQ
-- **Batch operations** — SendMessageBatch, DeleteMessageBatch, ChangeMessageVisibilityBatch with entry ID validation (`InvalidBatchEntryId`) and total batch size validation (`BatchRequestTooLong`)
-- **Queue attribute range validation** — validates `VisibilityTimeout`, `DelaySeconds`, `ReceiveMessageWaitTimeSeconds`, `MaximumMessageSize`, and `MessageRetentionPeriod` on both CreateQueue and SetQueueAttributes
-- **Message size validation** — rejects messages exceeding 1 MiB (1,048,576 bytes)
-- **Unicode character validation** — rejects messages with characters outside the AWS-allowed set
-- **KMS attributes** — `KmsMasterKeyId` and `KmsDataKeyReusePeriodSeconds` are accepted and stored (no actual encryption)
-- **FIFO queues** — `.fifo` suffix enforcement, `MessageGroupId` ordering, per-group locking (one inflight message per group), `MessageDeduplicationId`, content-based deduplication, sequence numbers, and FIFO-aware DLQ support
-- **Queue tags**
-
-## SNS Features
-
-- **SNS-to-SQS fan-out** — publish to a topic and messages are delivered to all confirmed SQS subscriptions
-- **Filter policies** — both `MessageAttributes` and `MessageBody` scope, supporting exact match, prefix, suffix, anything-but (including anything-but with suffix), numeric ranges, exists, null conditions, and `$or` top-level grouping. MessageBody scope supports nested key matching
-- **Raw message delivery** — configurable per subscription
-- **Message size validation** — rejects messages exceeding 256 KB (262,144 bytes)
-- **Topic idempotency with conflict detection** — `CreateTopic` returns the existing topic when called with the same name, attributes, and tags, but throws when attributes or tags differ
-- **Subscription idempotency with conflict detection** — `Subscribe` returns the existing subscription when the same (topic, protocol, endpoint) combination is used with matching attributes, but throws when attributes differ
-- **Subscription attribute validation** — `SetSubscriptionAttributes` validates attribute names and rejects unknown or read-only attributes
-- **Topic and subscription tags**
-- **FIFO topics** — `.fifo` suffix enforcement, `MessageGroupId` and `MessageDeduplicationId` passthrough to SQS subscriptions, content-based deduplication
-- **Batch publish**
-
-## S3 Features
-
-- **Bucket management** — CreateBucket (idempotent), DeleteBucket (rejects non-empty), HeadBucket, ListBuckets, ListObjects (V1 and V2)
-- **Object operations** — PutObject, GetObject, DeleteObject, HeadObject, CopyObject with ETag, Content-Type, and Last-Modified headers
-- **Multipart uploads** — CreateMultipartUpload, UploadPart, UploadPartCopy, CompleteMultipartUpload, AbortMultipartUpload with correct multipart ETag calculation (`MD5-of-part-digests-partCount`), metadata preservation, and part overwrite support
-- **ListObjects V2** — prefix filtering, delimiter-based virtual directories, MaxKeys, continuation tokens, StartAfter
-- **CopyObject** — same-bucket and cross-bucket copy via `x-amz-copy-source` header, with metadata preservation
-- **User metadata** — `x-amz-meta-*` headers are stored and returned on GetObject and HeadObject
-- **Bulk delete** — DeleteObjects for batch key deletion with proper XML entity handling
-- **Keys with slashes** — full support for slash-delimited keys (e.g., `path/to/file.txt`)
-- **Stream uploads** — handles AWS chunked transfer encoding (`Content-Encoding: aws-chunked`) for stream bodies, including trailing header parsing for checksums
-- **Checksums** — CRC32, SHA1, and SHA256 checksums are stored on upload (PutObject, UploadPart) and returned on download (GetObject, HeadObject with `x-amz-checksum-mode: ENABLED`). Multipart uploads compute composite checksums. GetObjectAttributes supports the `Checksum` attribute. CRC32C and CRC64NVME are silently ignored. Checksums are stored and returned as-is — no body validation is performed.
-- **GetObjectAttributes** — selective metadata retrieval via `x-amz-object-attributes` header: ETag, StorageClass, ObjectSize, ObjectParts (with pagination), and Checksum (including per-part checksums for multipart objects)
-- **Path-style and virtual-hosted-style** — both S3 URL styles are supported (see below)
-
-### S3 URL styles
-
-The AWS SDK sends S3 requests using virtual-hosted-style URLs by default (e.g., `my-bucket.s3.localhost:4566`). This requires `*.localhost` to resolve to `127.0.0.1`. fauxqs supports several approaches.
-
-#### Option 1: `fauxqs.dev` wildcard DNS (recommended for Docker image)
-
-Works out of the box when running the [official Docker image](#running-with-docker) — nothing to configure. The `fauxqs.dev` domain provides wildcard DNS — `*.localhost.fauxqs.dev` resolves to `127.0.0.1` via a public DNS entry. Just use `s3.localhost.fauxqs.dev` as your endpoint. This replicates the approach [pioneered by LocalStack](https://docs.localstack.cloud/aws/services/s3/) with `localhost.localstack.cloud`: a public DNS record maps all subdomains to localhost, so virtual-hosted-style requests work without `/etc/hosts` changes, custom request handlers, or `forcePathStyle`. Works from any language, `fetch()`, or CLI tool.
-
-```typescript
-import { S3Client } from "@aws-sdk/client-s3";
-
-const s3 = new S3Client({
-  endpoint: "http://s3.localhost.fauxqs.dev:4566",
-  region: "us-east-1",
-  credentials: { accessKeyId: "test", secretAccessKey: "test" },
-});
-```
-
-You can also use raw HTTP requests:
-
-```bash
-# Upload
-curl -X PUT --data-binary @file.txt http://my-bucket.s3.localhost.fauxqs.dev:4566/file.txt
-
-# Download
-curl http://my-bucket.s3.localhost.fauxqs.dev:4566/file.txt
-```
-
-This is the recommended approach for host-to-Docker setups. If you are using fauxqs as an [embedded library](#programmatic-usage) in Node.js tests, prefer Option 2 (`interceptLocalhostDns`) instead — it patches DNS globally so all clients work without modification, and requires no external DNS.
-
-For **container-to-container** S3 virtual-hosted-style in docker-compose, use the [built-in DNS server](#container-to-container-s3-virtual-hosted-style) instead — it resolves `*.s3.fauxqs` to the fauxqs container IP so other containers can use virtual-hosted-style S3 without `forcePathStyle`.
-
-#### Option 2: `interceptLocalhostDns()` (recommended for embedded library)
-
-Patches Node.js `dns.lookup` so that any hostname ending in `.localhost` resolves to `127.0.0.1`. No client changes needed.
-
-```typescript
-import { interceptLocalhostDns } from "fauxqs";
-
-const restore = interceptLocalhostDns();
-
-const s3 = new S3Client({
-  endpoint: "http://s3.localhost:4566",
-  region: "us-east-1",
-  credentials: { accessKeyId: "test", secretAccessKey: "test" },
-});
-
-// When done (e.g., in afterAll):
-restore();
-```
-
-The suffix is configurable: `interceptLocalhostDns("myhost.test")` matches `*.myhost.test`.
-
-**Tradeoffs:** Affects all DNS lookups in the process. Best suited for test suites (`beforeAll` / `afterAll`).
-
-#### Option 3: `createLocalhostHandler()` (per-client)
-
-Creates an HTTP request handler that resolves all hostnames to `127.0.0.1`. Scoped to a single client instance — no side effects, no external DNS dependency.
-
-```typescript
-import { S3Client } from "@aws-sdk/client-s3";
-import { createLocalhostHandler } from "fauxqs";
-
-const s3 = new S3Client({
-  endpoint: "http://s3.localhost:4566",
-  region: "us-east-1",
-  credentials: { accessKeyId: "test", secretAccessKey: "test" },
-  requestHandler: createLocalhostHandler(),
-});
-```
-
-#### Option 4: `forcePathStyle` (simplest fallback)
-
-Forces the SDK to use path-style URLs (`http://localhost:4566/my-bucket/key`) instead of virtual-hosted-style. No DNS or handler changes needed, but affects how the SDK resolves S3 URLs at runtime.
-
-```typescript
-const s3 = new S3Client({
-  endpoint: "http://localhost:4566",
-  forcePathStyle: true,
-  // ...
-});
-```
-
-
-### Using with AWS CLI
-
-fauxqs is wire-compatible with the standard AWS CLI. Point it at the fauxqs endpoint:
-
-#### SQS
-
-```bash
-aws --endpoint-url http://localhost:4566 sqs create-queue --queue-name my-queue
-aws --endpoint-url http://localhost:4566 sqs create-queue \
-  --queue-name my-queue.fifo \
-  --attributes FifoQueue=true,ContentBasedDeduplication=true
-aws --endpoint-url http://localhost:4566 sqs send-message \
-  --queue-url http://localhost:4566/000000000000/my-queue \
-  --message-body "hello"
-```
-
-#### SNS
-
-```bash
-aws --endpoint-url http://localhost:4566 sns create-topic --name my-topic
-aws --endpoint-url http://localhost:4566 sns subscribe \
-  --topic-arn arn:aws:sns:us-east-1:000000000000:my-topic \
-  --protocol sqs \
-  --notification-endpoint arn:aws:sqs:us-east-1:000000000000:my-queue
-```
-
-#### S3
-
-```bash
-aws --endpoint-url http://localhost:4566 s3 mb s3://my-bucket
-aws --endpoint-url http://localhost:4566 s3 cp file.txt s3://my-bucket/file.txt
-```
-
-If the AWS CLI uses virtual-hosted-style S3 URLs by default, configure path-style:
-
-```bash
-aws configure set default.s3.addressing_style path
-```
-
-## Testing Strategies
-
-fauxqs supports two deployment modes that complement each other for a complete testing workflow:
-
-| Mode | Best for | Startup | Assertions |
-|------|----------|---------|------------|
-| **Library** (embedded) | Unit tests, integration tests, CI | Milliseconds, in-process | Full programmatic API: `sendMessage`, `publish`, `spy`, `inspectQueue`, `reset`, `purgeAll` |
-| **Docker** (standalone) | Local development, acceptance tests, dev environments | Seconds, real HTTP | Init config, HTTP inspection endpoints |
-
-### Library mode for tests
-
-Embed fauxqs directly in your test suite. Each test file gets its own server instance on a random port — no Docker dependency, no shared state, no port conflicts:
-
-```typescript
-// test/setup.ts
-import { startFauxqs, type FauxqsServer } from "fauxqs";
-
-export async function createTestServer(): Promise<FauxqsServer> {
-  const server = await startFauxqs({ port: 0, logger: false, messageSpies: true });
-
-  // Pre-create resources via the programmatic API (no SDK roundtrips)
-  server.createQueue("my-queue");
-  server.createBucket("my-bucket");
-
-  return server;
-}
-```
-
-```typescript
-// test/app.test.ts
-import { describe, it, expect, beforeAll, afterAll, beforeEach } from "vitest";
-
-let server: FauxqsServer;
-
-beforeAll(async () => { server = await createTestServer(); });
-afterAll(async () => { await server.stop(); });
-beforeEach(() => { server.spy.clear(); });
-
-it("tracks uploads via the spy", async () => {
-  // ... trigger your app logic that uploads to S3 ...
-
-  const event = await server.spy.waitForMessage(
-    { service: "s3", bucket: "my-bucket", key: "file.txt", status: "uploaded" },
-    undefined,
-    2000, // timeout — prevents tests from hanging
-  );
-  expect(event.status).toBe("uploaded");
-});
-```
-
-Library mode gives you deterministic assertions via the [message spy](#message-spy), non-destructive state inspection via [`inspectQueue()`](#queue-inspection), and instant state reset with `reset()` / `purgeAll()` — none of which are available from outside the process.
-
-### Docker mode for local development
-
-Use `docker-compose.yml` with an init config to give your team a consistent local environment:
-
-```yaml
-# docker-compose.yml
-services:
-  fauxqs:
-    image: kibertoad/fauxqs:latest
-    ports: ["4566:4566"]
-    environment:
-      - FAUXQS_INIT=/app/init.json
-    volumes:
-      - ./fauxqs-init.json:/app/init.json
-
-  app:
-    build: .
-    depends_on:
-      fauxqs:
-        condition: service_healthy
-    environment:
-      - AWS_ENDPOINT=http://fauxqs:4566
-```
-
-Docker mode validates your real deployment topology — networking, DNS, container-to-container communication — and is language-agnostic (any AWS SDK can connect).
-
-### Recommended combination
-
-Use both modes together. Library mode runs in CI on every commit (fast, no Docker required). Docker mode runs locally via `docker compose up` and optionally in a separate CI stage for acceptance testing.
-
-See the [`examples/recommended/`](examples/recommended/) directory for a complete working example with a Fastify app, library-mode vitest tests, and Docker compose configuration.
-
-## Conventions
-
-- Account ID: `000000000000`
-- Region: auto-detected from SDK `Authorization` header; falls back to `defaultRegion` (defaults to `us-east-1`). Region is part of entity identity — same-name entities in different regions are independent.
-- Queue URL format: `http://sqs.{region}.{host}:{port}/000000000000/{queueName}` (host defaults to `localhost`)
-- Queue ARN format: `arn:aws:sqs:{region}:000000000000:{queueName}`
-- Topic ARN format: `arn:aws:sns:{region}:000000000000:{topicName}`
-
-## Limitations
-
-fauxqs is designed for development and testing. It does not support:
-
-- Non-SQS SNS delivery protocols (HTTP/S, Lambda, email, SMS)
-- Persistence across restarts
-- Authentication or authorization
-- Cross-account operations
-
-## Examples
-
-The [`examples/`](examples/) directory contains runnable TypeScript examples covering fauxqs-specific features beyond standard AWS SDK usage:
-
-| Example | Description |
-|---------|-------------|
-| [`alternatives/programmatic/programmatic-api.ts`](examples/alternatives/programmatic/programmatic-api.ts) | Server lifecycle, resource creation, SDK usage, `inspectQueue()`, `reset()`, `purgeAll()`, `setup()` |
-| [`alternatives/programmatic/message-spy.ts`](examples/alternatives/programmatic/message-spy.ts) | `MessageSpyReader` — all spy methods, partial/predicate filters, discriminated union narrowing, DLQ tracking |
-| [`alternatives/programmatic/init-config.ts`](examples/alternatives/programmatic/init-config.ts) | File-based and inline init config, DLQ chains, `setup()` idempotency, purge + re-apply pattern |
-| [`alternatives/programmatic/queue-inspection.ts`](examples/alternatives/programmatic/queue-inspection.ts) | Programmatic `inspectQueue()` and HTTP `/_fauxqs/queues` endpoints |
-| [`alternatives/docker/standalone/`](examples/alternatives/docker/standalone/standalone-container.ts) | Connecting to a fauxqs Docker container from the host |
-| [`alternatives/docker/container-to-container/`](examples/alternatives/docker/container-to-container/) | Container-to-container communication via docker-compose |
-| [`recommended/`](examples/recommended/) | Dual-mode testing: library mode (vitest + spy) for CI, Docker for local dev |
-
-All examples are type-checked in CI to prevent staleness.
-
-## Migrating from LocalStack
-
-If you're currently using LocalStack for local SNS, SQS, and/or S3 emulation, fauxqs is a drop-in replacement for those services. Both listen on port 4566 by default and accept the same AWS SDK calls, so the migration is straightforward.
-
-There are two approaches: a Docker swap (quickest) and a hybrid setup (recommended for the best integration test experience). Which one makes sense depends on whether you use S3.
-
-### SNS/SQS only
-
-If your LocalStack usage is limited to SNS and SQS, the migration is a one-line Docker image swap. No SDK client changes are needed — the endpoint URL, port, and credentials stay the same.
-
-**Docker swap:**
-
-```yaml
-# Before (LocalStack)
-services:
-  localstack:
-    image: localstack/localstack
-    ports: ["4566:4566"]
-    environment:
-      - SERVICES=sqs,sns
-
-# After (fauxqs)
-services:
-  fauxqs:
-    image: kibertoad/fauxqs:latest
-    ports: ["4566:4566"]
-```
-
-**Region:** Both LocalStack and fauxqs auto-detect the region from the SDK client's `Authorization` header, so in most setups no configuration is needed. If you have older LocalStack configs that used the now-removed `DEFAULT_REGION` env var (deprecated in v0.12.7, removed in v2.0), the equivalent in fauxqs is `FAUXQS_DEFAULT_REGION` — it serves as a fallback when the region can't be resolved from request headers. Both default to `us-east-1`.
-
-The main difference is how resources are pre-created. LocalStack uses init hooks (`/etc/localstack/init/ready.d/` shell scripts with `awslocal` CLI calls), while fauxqs uses a declarative JSON config. `awslocal` defaults to `us-east-1` unless you pass `--region`, and fauxqs init config uses `defaultRegion` (`us-east-1`) unless you set an explicit `region` per resource — so both create resources in the same region by default:
-
-```bash
-# LocalStack init script (ready.d/init.sh)
-# awslocal defaults to us-east-1; use --region to override
-awslocal sqs create-queue --queue-name orders
-awslocal sqs create-queue --queue-name orders-dlq
-awslocal sns create-topic --name events
-awslocal sns subscribe \
-  --topic-arn arn:aws:sns:us-east-1:000000000000:events \
-  --protocol sqs \
-  --notification-endpoint arn:aws:sqs:us-east-1:000000000000:orders
-```
-
-```json
-// fauxqs init.json — uses defaultRegion (us-east-1) unless "region" is set per resource
-{
-  "queues": [{ "name": "orders" }, { "name": "orders-dlq" }],
-  "topics": [{ "name": "events" }],
-  "subscriptions": [{ "topic": "events", "queue": "orders" }]
-}
-```
-
-```yaml
-# Before (LocalStack docker-compose.yml)
-services:
-  localstack:
-    image: localstack/localstack
-    ports: ["4566:4566"]
-    environment:
-      - SERVICES=sqs,sns
-    volumes:
-      - ./ready.d:/etc/localstack/init/ready.d
-
-# After (fauxqs docker-compose.yml)
-services:
-  fauxqs:
-    image: kibertoad/fauxqs:latest
-    ports: ["4566:4566"]
-    environment:
-      - FAUXQS_INIT=/app/init.json
-    volumes:
-      - ./init.json:/app/init.json
-```
-
-Update your docker-compose service name references (e.g., `http://localstack:4566` to `http://fauxqs:4566`) and you're done.
-
-### SNS/SQS/S3
-
-When S3 is involved, the Docker swap is still straightforward — the only additional consideration is S3 URL style. If you were using `forcePathStyle: true` with LocalStack, it works identically with fauxqs. If you were using LocalStack's `localhost.localstack.cloud` wildcard DNS for virtual-hosted-style, switch to `localhost.fauxqs.dev`:
-
-```typescript
-// Before (LocalStack)
-const s3 = new S3Client({
-  endpoint: "http://s3.localhost.localstack.cloud:4566",
-  // ...
-});
-
-// After (fauxqs)
-const s3 = new S3Client({
-  endpoint: "http://s3.localhost.fauxqs.dev:4566",
-  // ...
-});
-```
-
-For container-to-container S3 in docker-compose, fauxqs includes a built-in dnsmasq that resolves `*.s3.fauxqs` to the container IP — see [Container-to-container S3 virtual-hosted-style](#container-to-container-s3-virtual-hosted-style).
-
-The init config for S3 is the same declarative JSON, with a `buckets` array:
-
-```json
-{
-  "queues": [{ "name": "orders" }],
-  "topics": [{ "name": "events" }],
-  "subscriptions": [{ "topic": "events", "queue": "orders" }],
-  "buckets": ["uploads", "exports"]
-}
-```
-
-### Going hybrid (recommended)
-
-The Docker swap gets you running quickly, but the real win comes from going hybrid: use fauxqs as an **embedded library** in your test suite and keep Docker for local development.
-
-With LocalStack, integration tests typically look like this:
-
-1. Start LocalStack container (docker-compose or testcontainers) — takes seconds
-2. Create resources via `awslocal` or SDK calls — more seconds
-3. Run your test logic
-4. Assert by polling SQS queues, checking S3 objects, etc.
-5. Clean up resources between tests — often fragile or skipped
-
-With fauxqs in library mode:
-
-1. `startFauxqs({ port: 0 })` — starts in milliseconds, in-process
-2. `server.setup({ queues: [...], topics: [...] })` — instant, no network calls
-3. Run your test logic
-4. Assert with `server.spy.waitForMessage()` — no polling, no race conditions
-5. `server.reset()` between tests — clears messages, keeps resources
-
-**What you gain:**
-
-| Concern | LocalStack Docker | fauxqs library mode |
-|---------|-------------------|---------------------|
-| Test startup | Seconds (container boot + resource creation) | Milliseconds (in-process) |
-| CI dependency | Docker required | npm only |
-| Asserting message delivery | Poll SQS queue, hope timing is right | `spy.waitForMessage()` — resolves immediately or waits |
-| Asserting message *not* delivered | `sleep()` + check | `spy.expectNoMessage()` — deterministic negative assertion |
-| Filter policy testing | Receive from queue, check absence manually | `expectNoMessage()` on filtered-out queues |
-| DLQ verification | Receive from DLQ queue via SDK | `spy.waitForMessage({ status: "dlq" })` + `inspectQueue()` |
-| Queue state inspection | `GetQueueAttributes` (counts only) | `inspectQueue()` — see every message, grouped by state |
-| State reset between tests | Restart container or re-create resources | `server.reset()` — instant, preserves resource definitions |
-| Seeding test data | SDK calls through network stack | `server.sendMessage()` / `server.publish()` — direct, no network |
-| S3 event tracking | Check bucket contents via SDK | `spy.waitForMessage({ service: "s3", status: "uploaded" })` |
-| Parallel test files | Port conflicts or shared state | Each file gets its own server on port 0 |
-
-**Migration path:**
-
-1. Install fauxqs as a dev dependency: `npm install -D fauxqs`
-2. Create a test helper:
-
-```typescript
-// test/setup.ts
-import { startFauxqs, type FauxqsServer } from "fauxqs";
-
-export async function createTestServer(): Promise<FauxqsServer> {
-  const server = await startFauxqs({ port: 0, logger: false, messageSpies: true });
-
-  server.setup({
-    queues: [{ name: "orders" }, { name: "orders-dlq" }],
-    topics: [{ name: "events" }],
-    subscriptions: [{ topic: "events", queue: "orders" }],
-    buckets: ["uploads"],
-  });
-
-  return server;
-}
-```
-
-3. Replace your LocalStack container setup with the test helper:
-
-```typescript
-// Before: LocalStack via testcontainers or docker-compose
-let endpoint: string;
-beforeAll(async () => {
-  // start container, wait for health, create resources via SDK...
-  endpoint = "http://localhost:4566";
-}, 30_000);
-
-// After: fauxqs library
-let server: FauxqsServer;
-beforeAll(async () => {
-  server = await createTestServer();
-});
-afterAll(async () => { await server.stop(); });
-beforeEach(() => { server.reset(); });
-```
-
-4. Replace polling-based assertions with spy-based ones:
-
-```typescript
-// Before: poll and hope
-await sqsClient.send(new SendMessageCommand({ QueueUrl: queueUrl, MessageBody: "test" }));
-const result = await sqsClient.send(new ReceiveMessageCommand({ QueueUrl: queueUrl, WaitTimeSeconds: 5 }));
-expect(result.Messages?.[0]?.Body).toBe("test");
-
-// After: spy knows immediately
-await sqsClient.send(new SendMessageCommand({ QueueUrl: queueUrl, MessageBody: "test" }));
-const msg = await server.spy.waitForMessage(
-  { service: "sqs", queueName: "orders", status: "published" },
-  undefined,
-  2000,
-);
-expect(msg.body).toBe("test");
-```
-
-5. Keep your `docker-compose.yml` with the fauxqs image for local development — `docker compose up` gives your team a running environment without Node.js installed.
-
-This hybrid setup gives you fast, deterministic tests in CI (no Docker required) and a realistic Docker environment for local development. See the [`examples/recommended/`](examples/recommended/) directory for a complete working example.
-
-## Benchmarks
-
-SQS throughput benchmarks are available in the [`benchmarks/`](benchmarks/) directory, comparing fauxqs across different deployment modes (in-process library, official Docker image, lightweight Docker container) and against LocalStack. See [`benchmarks/BENCHMARKING.md`](benchmarks/BENCHMARKING.md) for setup descriptions, instructions, and how to interpret results.
-
-## License
-
-MIT
+<img src="logo-readme.jpg" alt="fauxqs" width="360" />
+
+[![npm version](https://img.shields.io/npm/v/fauxqs.svg)](https://www.npmjs.com/package/fauxqs)
+
+# fauxqs
+
+Local SNS/SQS/S3 emulator for development and testing. Point your `@aws-sdk/client-sqs`, `@aws-sdk/client-sns`, and `@aws-sdk/client-s3` clients at fauxqs instead of real AWS.
+
+All state is in-memory by default. Optional SQLite-based persistence is available via the `dataDir` option.
+
+## Table of Contents
+
+- [Installation](#installation)
+- [Usage](#usage)
+  - [Running the server](#running-the-server)
+  - [Running in the background](#running-in-the-background)
+  - [Running with Docker](#running-with-docker)
+  - [Running in Docker Compose](#running-in-docker-compose)
+    - [Container-to-container S3 virtual-hosted-style](#container-to-container-s3-virtual-hosted-style)
+  - [Configuring AWS SDK clients](#configuring-aws-sdk-clients)
+  - [Programmatic usage](#programmatic-usage)
+    - [Relaxed rules](#relaxed-rules)
+    - [Programmatic state setup](#programmatic-state-setup)
+    - [Sending messages programmatically](#sending-messages-programmatically)
+    - [Init config file](#init-config-file)
+    - [Init config schema reference](#init-config-schema-reference)
+    - [Message spy](#message-spy)
+    - [Queue inspection](#queue-inspection)
+  - [Persistence](#persistence)
+  - [Configurable queue URL host](#configurable-queue-url-host)
+  - [Region](#region)
+- [Supported API Actions](#supported-api-actions)
+  - [SQS](#sqs)
+  - [SNS](#sns)
+  - [S3](#s3)
+  - [STS](#sts)
+- [SQS Features](#sqs-features)
+- [SNS Features](#sns-features)
+- [S3 Features](#s3-features)
+  - [S3 URL styles](#s3-url-styles)
+  - [Using with AWS CLI](#using-with-aws-cli)
+- [Testing Strategies](#testing-strategies)
+  - [Library mode for tests](#library-mode-for-tests)
+  - [Docker mode for local development](#docker-mode-for-local-development)
+  - [Recommended combination](#recommended-combination)
+- [Conventions](#conventions)
+- [Limitations](#limitations)
+- [Examples](#examples)
+- [Migrating from LocalStack](#migrating-from-localstack)
+  - [SNS/SQS only](#snssqs-only)
+  - [SNS/SQS/S3](#snssqss3)
+  - [Going hybrid (recommended)](#going-hybrid-recommended)
+- [Benchmarks](#benchmarks)
+- [License](#license)
+
+## Installation
+
+**Docker** (recommended for standalone usage) — [Docker Hub](https://hub.docker.com/r/kibertoad/fauxqs):
+
+```bash
+docker run -p 4566:4566 kibertoad/fauxqs
+```
+
+**npm** (for embedded library usage or CLI):
+
+```bash
+npm install fauxqs
+```
+
+> **Node.js compatibility:** fauxqs requires Node.js 22.5+ (for `node:sqlite`) when used as a library. If your project is on Node.js 20, you can either use `fauxqs@1.13.0` (last version with Node.js 20 support) or run the latest fauxqs as a [Docker container](#running-with-docker).
+
+## Usage
+
+### Running the server
+
+```bash
+npx fauxqs
+```
+
+The server starts on port `4566` and handles SQS, SNS, and S3 on a single endpoint.
+
+#### Environment variables
+
+| Variable | Description | Default |
+|----------|-------------|---------|
+| `FAUXQS_PORT` | Port to listen on | `4566` |
+| `FAUXQS_HOST` | Host for queue URLs (`sqs.<region>.<host>` format) | `localhost` |
+| `FAUXQS_DEFAULT_REGION` | Fallback region for ARNs and URLs | `us-east-1` |
+| `FAUXQS_LOGGER` | Enable request logging (`true`/`false`) | `true` |
+| `FAUXQS_INIT` | Path to a JSON init config file (see [Init config file](#init-config-file)) | (none) |
+| `FAUXQS_DATA_DIR` | Directory for SQLite persistence (see [Persistence](#persistence)). Omit to keep all state in-memory. | (none) |
+| `FAUXQS_PERSISTENCE` | Set to `true` to enable persistence when `FAUXQS_DATA_DIR` is set | `false` |
+| `FAUXQS_DNS_NAME` | Domain that dnsmasq resolves (including all subdomains) to the container IP. Only needed when the container hostname doesn't match the docker-compose service name — e.g., when using `container_name` or running with plain `docker run`. In docker-compose the hostname is set to the service name automatically, so this is rarely needed. (Docker only) | container hostname |
+| `FAUXQS_DNS_UPSTREAM` | Where dnsmasq forwards non-fauxqs DNS queries (e.g., `registry.npmjs.org`). Change this if you're in a corporate network with an internal DNS server, or if you prefer a different public resolver like `1.1.1.1`. (Docker only) | `8.8.8.8` |
+
+```bash
+FAUXQS_PORT=3000 FAUXQS_INIT=init.json npx fauxqs
+```
+
+A health check is available at `GET /health`.
+
+### Running in the background
+
+To keep fauxqs running while you work on your app or run tests repeatedly, start it as a background process:
+
+```bash
+npx fauxqs &
+```
+
+Or in a separate terminal:
+
+```bash
+npx fauxqs
+```
+
+All state accumulates in memory across requests, so queues, topics, and objects persist until the server is stopped.
+
+To stop the server:
+
+```bash
+# If backgrounded in the same shell
+kill %1
+
+# Cross-platform, by port
+npx cross-port-killer 4566
+```
+
+### Running with Docker
+
+The official Docker image is available on Docker Hub:
+
+```bash
+docker run -p 4566:4566 kibertoad/fauxqs
+```
+
+To persist state across container restarts, mount a volume at `/data` and set `FAUXQS_PERSISTENCE=true`:
+
+```bash
+docker run -p 4566:4566 -v fauxqs-data:/data -e FAUXQS_PERSISTENCE=true kibertoad/fauxqs
+```
+
+With an init config file:
+
+```bash
+docker run -p 4566:4566 \
+  -v fauxqs-data:/data \
+  -v ./init.json:/app/init.json \
+  -e FAUXQS_INIT=/app/init.json \
+  kibertoad/fauxqs
+```
+
+### Running in Docker Compose
+
+Use the `kibertoad/fauxqs` image and mount a JSON init config to pre-create resources on startup:
+
+```json
+// scripts/fauxqs/init.json
+{
+  "queues": [
+    {
+      "name": "my-queue.fifo",
+      "attributes": { "FifoQueue": "true", "ContentBasedDeduplication": "true" }
+    },
+    { "name": "my-dlq" }
+  ],
+  "topics": [{ "name": "my-events" }],
+  "subscriptions": [{ "topic": "my-events", "queue": "my-dlq" }],
+  "buckets": ["my-uploads"]
+}
+```
+
+```yaml
+# docker-compose.yml
+services:
+  fauxqs:
+    image: kibertoad/fauxqs:latest
+    ports:
+      - "4566:4566"
+    environment:
+      - FAUXQS_INIT=/app/init.json
+    volumes:
+      - ./scripts/fauxqs/init.json:/app/init.json
+      - fauxqs-data:/data
+
+  app:
+    # ...
+    depends_on:
+      fauxqs:
+        condition: service_healthy
+
+volumes:
+  fauxqs-data:
+```
+
+The image has a built-in `HEALTHCHECK`, so `service_healthy` works without extra configuration in your compose file. Other containers reference fauxqs using the Docker service name (`http://fauxqs:4566`). The init config file creates all queues, topics, subscriptions, and buckets before the healthcheck passes, so dependent services start only after resources are ready. The `fauxqs-data` volume persists state across `docker compose down` / `up` cycles — queues, messages, objects, and all other state are restored on startup. Init config is idempotent, so re-applying it after a restart skips resources that already exist.
+
+#### Container-to-container S3 virtual-hosted-style
+
+The Docker image includes a built-in DNS server ([dnsmasq](https://thekelleys.org.uk/dnsmasq/doc.html)) that resolves the container hostname and all its subdomains (e.g., `fauxqs`, `s3.fauxqs`, `my-bucket.s3.fauxqs`) to the container's own IP. This enables virtual-hosted-style S3 from other containers without `forcePathStyle`.
+
+To use it, assign fauxqs a static IP and point other containers' DNS to it:
+
+```yaml
+# docker-compose.yml
+services:
+  fauxqs:
+    image: kibertoad/fauxqs:latest
+    networks:
+      default:
+        ipv4_address: 10.0.0.2
+    ports:
+      - "4566:4566"
+    environment:
+      - FAUXQS_INIT=/app/init.json
+      - FAUXQS_HOST=fauxqs
+    volumes:
+      - ./scripts/fauxqs/init.json:/app/init.json
+      - fauxqs-data:/data
+
+  app:
+    dns: 10.0.0.2
+    depends_on:
+      fauxqs:
+        condition: service_healthy
+    environment:
+      - AWS_ENDPOINT=http://s3.fauxqs:4566
+
+volumes:
+  fauxqs-data:
+
+networks:
+  default:
+    ipam:
+      config:
+        - subnet: 10.0.0.0/24
+```
+
+From the `app` container, `my-bucket.s3.fauxqs` resolves to `10.0.0.2` (the fauxqs container), so virtual-hosted-style S3 works:
+
+```typescript
+const s3 = new S3Client({
+  endpoint: "http://s3.fauxqs:4566",
+  region: "us-east-1",
+  credentials: { accessKeyId: "test", secretAccessKey: "test" },
+  // No forcePathStyle needed!
+});
+```
+
+The DNS server is configured automatically using the container hostname (which docker-compose sets to the service name), so in most setups no extra configuration is needed. See the [environment variables table](#environment-variables) for `FAUXQS_DNS_NAME` and `FAUXQS_DNS_UPSTREAM` if you need to override the defaults.
+
+### Configuring AWS SDK clients
+
+Point your SDK clients at the local server:
+
+```typescript
+import { SQSClient } from "@aws-sdk/client-sqs";
+import { SNSClient } from "@aws-sdk/client-sns";
+import { S3Client } from "@aws-sdk/client-s3";
+
+const sqsClient = new SQSClient({
+  endpoint: "http://localhost:4566",
+  region: "us-east-1",
+  credentials: { accessKeyId: "test", secretAccessKey: "test" },
+});
+
+const snsClient = new SNSClient({
+  endpoint: "http://localhost:4566",
+  region: "us-east-1",
+  credentials: { accessKeyId: "test", secretAccessKey: "test" },
+});
+
+// Using fauxqs.dev wildcard DNS — no helpers or forcePathStyle needed
+const s3Client = new S3Client({
+  endpoint: "http://s3.localhost.fauxqs.dev:4566",
+  region: "us-east-1",
+  credentials: { accessKeyId: "test", secretAccessKey: "test" },
+});
+```
+
+Any credentials are accepted and never validated.
+
+> **Note:** The `fauxqs.dev` wildcard DNS (`*.localhost.fauxqs.dev` → `127.0.0.1`) replicates the approach [pioneered by LocalStack](https://hashnode.localstack.cloud/efficient-localstack-s3-endpoint-configuration) with `localhost.localstack.cloud`. A public DNS entry resolves all subdomains to localhost, so virtual-hosted-style S3 requests work without `/etc/hosts` changes, custom request handlers, or `forcePathStyle`. See [S3 URL styles](#s3-url-styles) for alternative approaches.
+
+### Programmatic usage
+
+You can also embed fauxqs directly in your test suite:
+
+```typescript
+import { startFauxqs } from "fauxqs";
+
+const server = await startFauxqs({ port: 4566, logger: false });
+
+console.log(server.address); // "http://127.0.0.1:4566"
+console.log(server.port);    // 4566
+
+// point your SDK clients at server.address
+
+// clean up when done
+await server.stop();
+```
+
+Pass `port: 0` to let the OS assign a random available port (useful in tests).
+
+#### Relaxed rules
+
+By default, fauxqs enforces AWS-strict validations. You can selectively relax some of these for convenience during development:
+
+```typescript
+const server = await startFauxqs({
+  port: 0,
+  relaxedRules: {
+    disableMinCopySourceSize: true,
+  },
+});
+```
+
+| Rule | Default | Description |
+|------|---------|-------------|
+| `disableMinCopySourceSize` | `false` | AWS requires the source object to be [larger than 5 MiB](https://docs.aws.amazon.com/AmazonS3/latest/API/API_UploadPartCopy.html) for byte-range `UploadPartCopy`. Set to `true` to allow byte-range copies from smaller sources. |
+
+#### Programmatic state setup
+
+The server object exposes methods for pre-creating resources without going through the SDK:
+
+```typescript
+const server = await startFauxqs({ port: 0, logger: false });
+
+// Create individual resources — each returns metadata about the created resource
+const { queueUrl, queueArn, queueName } = server.createQueue("my-queue");
+const { queueUrl: dlqUrl } = server.createQueue("my-dlq", {
+  attributes: { VisibilityTimeout: "60" },
+  tags: { env: "test" },
+});
+const { topicArn } = server.createTopic("my-topic");
+server.subscribe({ topic: "my-topic", queue: "my-queue" });
+const { bucketName } = server.createBucket("my-bucket");
+
+// Create resources in a specific region
+server.createQueue("eu-queue", { region: "eu-west-1" });
+server.createTopic("eu-topic", { region: "eu-west-1" });
+server.subscribe({ topic: "eu-topic", queue: "eu-queue", region: "eu-west-1" });
+
+// Or create everything at once
+server.setup({
+  queues: [
+    { name: "orders" },
+    { name: "notifications", attributes: { DelaySeconds: "5" } },
+    { name: "eu-orders", region: "eu-west-1" },
+  ],
+  topics: [{ name: "events" }],
+  subscriptions: [
+    { topic: "events", queue: "orders" },
+    { topic: "events", queue: "notifications" },
+  ],
+  buckets: ["uploads", "exports"],
+});
+
+// Delete individual resources (uses defaultRegion; pass { region } to override)
+server.deleteQueue("my-queue");                          // no-op if queue doesn't exist
+server.deleteQueue("eu-queue", { region: "eu-west-1" }); // delete in specific region
+server.deleteTopic("my-topic");                          // also removes associated subscriptions
+server.deleteTopic("eu-topic", { region: "eu-west-1" });
+server.emptyBucket("my-bucket");                         // removes all objects, keeps the bucket
+
+// Clear all messages and S3 objects between tests (keeps queues, topics, subscriptions, buckets)
+server.reset();
+
+// Or nuke everything — removes queues, topics, subscriptions, and buckets too
+server.purgeAll();
+```
+
+#### Sending messages programmatically
+
+Send SQS messages and publish to SNS topics without instantiating SDK clients:
+
+```typescript
+// SQS: enqueue a message directly
+const { messageId, md5OfBody } = server.sendMessage("my-queue", "hello world");
+
+// With message attributes
+server.sendMessage("my-queue", JSON.stringify({ orderId: "123" }), {
+  messageAttributes: {
+    eventType: { DataType: "String", StringValue: "order.created" },
+  },
+});
+
+// With delay
+server.sendMessage("my-queue", "delayed message", { delaySeconds: 10 });
+
+// FIFO queue — returns sequenceNumber
+const { sequenceNumber } = server.sendMessage("my-queue.fifo", "fifo message", {
+  messageGroupId: "group-1",
+  messageDeduplicationId: "dedup-1",
+});
+
+// SNS: publish to a topic (fans out to all SQS subscriptions)
+const { messageId: snsMessageId } = server.publish("my-topic", "event payload");
+
+// With subject and message attributes
+server.publish("my-topic", JSON.stringify({ orderId: "456" }), {
+  subject: "Order Update",
+  messageAttributes: {
+    eventType: { DataType: "String", StringValue: "order.updated" },
+  },
+});
+
+// FIFO topic
+server.publish("my-topic.fifo", "fifo event", {
+  messageGroupId: "group-1",
+  messageDeduplicationId: "dedup-1",
+});
+```
+
+`sendMessage` validates the message body (invalid characters, size limits), applies queue-level `DelaySeconds` defaults, handles FIFO deduplication, and emits spy events automatically. Returns `{ messageId, md5OfBody, md5OfMessageAttributes?, sequenceNumber? }`.
+
+`publish` validates message size, evaluates filter policies on each subscription, supports raw message delivery, and emits spy events. Returns `{ messageId }`.
+
+Both methods throw if the target queue or topic doesn't exist.
+
+#### Init config file
+
+Create a JSON file to pre-create resources on startup. The file is validated on load — malformed configs produce a clear error instead of silent failures.
+
+```json
+{
+  "queues": [
+    { "name": "orders" },
+    { "name": "orders-dlq" },
+    { "name": "orders.fifo", "attributes": { "FifoQueue": "true", "ContentBasedDeduplication": "true" } }
+  ],
+  "topics": [
+    { "name": "events" }
+  ],
+  "subscriptions": [
+    { "topic": "events", "queue": "orders" }
+  ],
+  "buckets": ["uploads", "exports"]
+}
+```
+
+Pass it via the `FAUXQS_INIT` environment variable or the `init` option:
+
+```bash
+FAUXQS_INIT=init.json npx fauxqs
+```
+
+```typescript
+const server = await startFauxqs({ init: "init.json" });
+// or inline:
+const server = await startFauxqs({
+  init: { queues: [{ name: "my-queue" }], buckets: ["my-bucket"] },
+});
+```
+
+#### Init config schema reference
+
+All top-level fields are optional. Resources are created in dependency order: queues, topics, subscriptions, buckets.
+
+##### `queues`
+
+Array of queue objects.
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| `name` | `string` | Yes | Queue name. Use `.fifo` suffix for FIFO queues. |
+| `region` | `string` | No | Override the default region for this queue. The queue's ARN and URL will use this region. |
+| `attributes` | `Record<string, string>` | No | Queue attributes (see table below). |
+| `tags` | `Record<string, string>` | No | Key-value tags for the queue. |
+
+Supported queue attributes:
+
+| Attribute | Default | Range / Values |
+|-----------|---------|----------------|
+| `VisibilityTimeout` | `"30"` | `0` – `43200` (seconds) |
+| `DelaySeconds` | `"0"` | `0` – `900` (seconds) |
+| `MaximumMessageSize` | `"1048576"` | `1024` – `1048576` (bytes) |
+| `MessageRetentionPeriod` | `"345600"` | `60` – `1209600` (seconds) |
+| `ReceiveMessageWaitTimeSeconds` | `"0"` | `0` – `20` (seconds) |
+| `RedrivePolicy` | — | JSON string: `{"deadLetterTargetArn": "arn:...", "maxReceiveCount": "5"}` |
+| `Policy` | — | Queue policy JSON string (stored, not enforced) |
+| `KmsMasterKeyId` | — | KMS key ID (stored, no actual encryption) |
+| `KmsDataKeyReusePeriodSeconds` | — | KMS data key reuse period (stored, no actual encryption) |
+| `FifoQueue` | — | `"true"` for FIFO queues (queue name must end with `.fifo`) |
+| `ContentBasedDeduplication` | — | `"true"` or `"false"` (FIFO queues only) |
+
+Example:
+
+```json
+{
+  "queues": [
+    {
+      "name": "orders",
+      "attributes": { "VisibilityTimeout": "60", "DelaySeconds": "5" },
+      "tags": { "env": "staging", "team": "platform" }
+    },
+    {
+      "name": "orders-dlq"
+    },
+    {
+      "name": "orders.fifo",
+      "attributes": {
+        "FifoQueue": "true",
+        "ContentBasedDeduplication": "true"
+      }
+    },
+    {
+      "name": "retry-queue",
+      "attributes": {
+        "RedrivePolicy": "{\"deadLetterTargetArn\":\"arn:aws:sqs:us-east-1:000000000000:orders-dlq\",\"maxReceiveCount\":\"3\"}"
+      }
+    },
+    {
+      "name": "eu-orders",
+      "region": "eu-west-1"
+    }
+  ]
+}
+```
+
+##### `topics`
+
+Array of topic objects.
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| `name` | `string` | Yes | Topic name. Use `.fifo` suffix for FIFO topics. |
+| `region` | `string` | No | Override the default region for this topic. The topic's ARN will use this region. |
+| `attributes` | `Record<string, string>` | No | Topic attributes (e.g., `DisplayName`). |
+| `tags` | `Record<string, string>` | No | Key-value tags for the topic. |
+
+Example:
+
+```json
+{
+  "topics": [
+    {
+      "name": "events",
+      "attributes": { "DisplayName": "Application Events" },
+      "tags": { "env": "staging" }
+    },
+    {
+      "name": "events.fifo",
+      "attributes": { "FifoQueue": "true", "ContentBasedDeduplication": "true" }
+    }
+  ]
+}
+```
+
+##### `subscriptions`
+
+Array of subscription objects. Referenced topics and queues must be defined in the same config (or already exist on the server).
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| `topic` | `string` | Yes | Topic name (not ARN) to subscribe to. |
+| `queue` | `string` | Yes | Queue name (not ARN) to deliver messages to. |
+| `region` | `string` | No | Override the default region. The topic and queue ARNs will be resolved in this region. |
+| `attributes` | `Record<string, string>` | No | Subscription attributes (see table below). |
+
+Supported subscription attributes:
+
+| Attribute | Values | Description |
+|-----------|--------|-------------|
+| `RawMessageDelivery` | `"true"` / `"false"` | Deliver the raw message body instead of the SNS envelope JSON. |
+| `FilterPolicy` | JSON string | SNS filter policy for message filtering (e.g., `"{\"color\": [\"blue\"]}"`) |
+| `FilterPolicyScope` | `"MessageAttributes"` / `"MessageBody"` | Whether the filter policy applies to message attributes or body. Defaults to `MessageAttributes`. |
+| `RedrivePolicy` | JSON string | Subscription-level dead-letter queue config. |
+| `DeliveryPolicy` | JSON string | Delivery retry policy (stored, not enforced). |
+| `SubscriptionRoleArn` | ARN string | IAM role ARN for delivery (stored, not enforced). |
+
+Example:
+
+```json
+{
+  "subscriptions": [
+    {
+      "topic": "events",
+      "queue": "orders",
+      "attributes": {
+        "RawMessageDelivery": "true",
+        "FilterPolicy": "{\"eventType\": [\"order.created\", \"order.updated\"]}"
+      }
+    },
+    {
+      "topic": "events",
+      "queue": "notifications"
+    }
+  ]
+}
+```
+
+##### `buckets`
+
+Array of bucket name strings or objects. Use the object form to create directory buckets (S3 Express One Zone).
+
+```json
+{
+  "buckets": [
+    "uploads",
+    "exports",
+    { "name": "my-directory-bucket", "type": "directory" }
+  ]
+}
+```
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| `name` | `string` | Yes | Bucket name. |
+| `type` | `"general-purpose"` \| `"directory"` | No | Bucket type. Defaults to `"general-purpose"`. Directory buckets support `RenameObject`. |
+
+#### Message spy
+
+`MessageSpyReader` lets you await specific events flowing through SQS, SNS, and S3 in your tests — without polling queues yourself. Inspired by `HandlerSpy` from `message-queue-toolkit`.
+
+Enable it with the `messageSpies` option:
+
+```typescript
+const server = await startFauxqs({ port: 0, logger: false, messageSpies: true });
+```
+
+The spy tracks events across all three services using a discriminated union on `service`:
+
+**SQS events** (`service: 'sqs'`):
+- **`published`** — message was enqueued (via SendMessage, SendMessageBatch, or SNS fan-out)
+- **`consumed`** — message was deleted (via DeleteMessage / DeleteMessageBatch)
+- **`dlq`** — message exceeded `maxReceiveCount` and was moved to a dead-letter queue
+
+**SNS events** (`service: 'sns'`):
+- **`published`** — message was published to a topic (before fan-out to SQS subscriptions)
+
+**S3 events** (`service: 's3'`):
+- **`uploaded`** — object was put (PutObject or CompleteMultipartUpload)
+- **`downloaded`** — object was retrieved (GetObject)
+- **`deleted`** — object was deleted (DeleteObject, only when key existed)
+- **`copied`** — object was copied (CopyObject; also emits `uploaded` for the destination)
+- **`renamed`** — object was renamed (RenameObject, directory buckets only)
+
+##### Awaiting messages
+
+```typescript
+// Wait for a specific SQS message (resolves immediately if already in buffer)
+const msg = await server.spy.waitForMessage(
+  (m) => m.service === "sqs" && m.body === "order.created" && m.queueName === "orders",
+  "published",
+);
+
+// Wait by SQS message ID
+const msg = await server.spy.waitForMessageWithId(messageId, "consumed");
+
+// Partial object match (deep-equal on specified fields)
+const msg = await server.spy.waitForMessage({ service: "sqs", queueName: "orders", status: "published" });
+
+// Wait for an SNS publish event
+const msg = await server.spy.waitForMessage({ service: "sns", topicName: "my-topic", status: "published" });
+
+// Wait for an S3 upload event
+const msg = await server.spy.waitForMessage({ service: "s3", bucket: "my-bucket", key: "file.txt", status: "uploaded" });
+```
+
+`waitForMessage` checks the buffer first (retroactive resolution). If no match is found, it returns a Promise that resolves when a matching message arrives.
+
+##### Timeout
+
+All `waitForMessage` and `waitForMessageWithId` calls accept an optional `timeout` parameter (ms) as the third argument. If no matching message arrives in time, the promise rejects with a timeout error — preventing tests from hanging indefinitely:
+
+```typescript
+// Reject after 2 seconds if no match
+const msg = await server.spy.waitForMessage(
+  { service: "sqs", queueName: "orders" },
+  "published",
+  2000,
+);
+
+// Also works with waitForMessageWithId
+const msg = await server.spy.waitForMessageWithId(messageId, "consumed", 5000);
+```
+
+##### Waiting for multiple messages
+
+`waitForMessages` collects `count` matching messages before resolving. It checks the buffer first, then awaits future arrivals:
+
+```typescript
+// Wait for 3 messages on the orders queue
+const msgs = await server.spy.waitForMessages(
+  { service: "sqs", queueName: "orders" },
+  { count: 3, status: "published", timeout: 5000 },
+);
+// msgs.length === 3
+```
+
+If the timeout expires before enough messages arrive, the promise rejects with a message showing how many were collected (e.g., `"collected 1/3"`).
+
+##### Negative assertions
+
+`expectNoMessage` asserts that no matching message appears within a time window. Useful for verifying that filter policies dropped a message or that a side effect did not occur:
+
+```typescript
+// Assert no message was delivered to the wrong queue (waits 200ms by default)
+await server.spy.expectNoMessage({ service: "sqs", queueName: "wrong-queue" });
+
+// Custom window and status filter
+await server.spy.expectNoMessage(
+  { service: "sqs", queueName: "orders" },
+  { status: "dlq", within: 500 },
+);
+```
+
+If a matching message is already in the buffer, `expectNoMessage` rejects immediately. If one arrives during the wait, it rejects with `"matching message arrived during wait"`.
+
+##### Synchronous check
+
+```typescript
+const msg = server.spy.checkForMessage(
+  (m) => m.service === "sqs" && m.queueName === "my-queue",
+  "published",
+);
+// returns SpyMessage | undefined
+```
+
+##### Buffer management
+
+```typescript
+// Get all tracked messages (oldest to newest)
+const all = server.spy.getAllMessages();
+
+// Clear buffer and reject pending waiters
+server.spy.clear();
+```
+
+The buffer defaults to 100 messages (FIFO eviction). Configure with:
+
+```typescript
+const server = await startFauxqs({
+  messageSpies: { bufferSize: 500 },
+});
+```
+
+##### Types
+
+`server.spy` returns a `MessageSpyReader` — a read-only interface that exposes query and await methods but not internal mutation (e.g. recording new events):
+
+```typescript
+interface MessageSpyReader {
+  waitForMessage(filter: MessageSpyFilter, status?: string, timeout?: number): Promise<SpyMessage>;
+  waitForMessageWithId(messageId: string, status?: string, timeout?: number): Promise<SpyMessage>;
+  waitForMessages(filter: MessageSpyFilter, options: WaitForMessagesOptions): Promise<SpyMessage[]>;
+  expectNoMessage(filter: MessageSpyFilter, options?: ExpectNoMessageOptions): Promise<void>;
+  checkForMessage(filter: MessageSpyFilter, status?: string): SpyMessage | undefined;
+  getAllMessages(): SpyMessage[];
+  clear(): void;
+}
+
+interface WaitForMessagesOptions {
+  count: number;
+  status?: string;
+  timeout?: number;
+}
+
+interface ExpectNoMessageOptions {
+  status?: string;
+  within?: number; // ms, defaults to 200
+}
+```
+
+`SpyMessage` is a discriminated union:
+
+```typescript
+interface SqsSpyMessage {
+  service: "sqs";
+  queueName: string;
+  messageId: string;
+  body: string;
+  messageAttributes: Record<string, MessageAttributeValue>;
+  status: "published" | "consumed" | "dlq";
+  timestamp: number;
+}
+
+interface SnsSpyMessage {
+  service: "sns";
+  topicArn: string;
+  topicName: string;
+  messageId: string;
+  body: string;
+  messageAttributes: Record<string, MessageAttributeValue>;
+  status: "published";
+  timestamp: number;
+}
+
+interface S3SpyEvent {
+  service: "s3";
+  bucket: string;
+  key: string;
+  status: "uploaded" | "downloaded" | "deleted" | "copied" | "renamed";
+  timestamp: number;
+}
+
+type SpyMessage = SqsSpyMessage | SnsSpyMessage | S3SpyEvent;
+```
+
+##### Spy disabled by default
+
+Accessing `server.spy` when `messageSpies` is not set throws an error. There is no overhead on the message flow when spies are disabled.
+
+#### Queue inspection
+
+Non-destructive inspection of SQS queue state — see all messages (ready, in-flight, and delayed) without consuming them or affecting visibility timeouts.
+
+##### Programmatic API
+
+```typescript
+const result = server.inspectQueue("my-queue");
+// result is undefined if queue doesn't exist
+if (result) {
+  console.log(result.name);           // "my-queue"
+  console.log(result.url);            // "http://sqs.us-east-1.localhost:4566/000000000000/my-queue"
+  console.log(result.arn);            // "arn:aws:sqs:us-east-1:000000000000:my-queue"
+  console.log(result.attributes);     // { VisibilityTimeout: "30", ... }
+  console.log(result.messages.ready);    // messages available for receive
+  console.log(result.messages.delayed);  // messages waiting for delay to expire
+  console.log(result.messages.inflight); // received but not yet deleted
+  // Each inflight entry includes: { message, receiptHandle, visibilityDeadline }
+}
+```
+
+##### HTTP endpoints
+
+```bash
+# List all queues with summary counts
+curl http://localhost:4566/_fauxqs/queues
+# [{ "name": "my-queue", "approximateMessageCount": 5, "approximateInflightCount": 2, "approximateDelayedCount": 0, ... }]
+
+# Inspect a specific queue (full state)
+curl http://localhost:4566/_fauxqs/queues/my-queue
+# { "name": "my-queue", "messages": { "ready": [...], "delayed": [...], "inflight": [...] }, ... }
+```
+
+Returns 404 for non-existent queues. Inspection never modifies queue state — messages remain exactly where they are.
+
+### Persistence
+
+By default (when using `npx fauxqs` or the programmatic API without `dataDir`), all state is in-memory and lost when the server stops. To persist state across restarts, set a `dataDir` — fauxqs will store all queues, messages, topics, subscriptions, buckets, and objects in a SQLite database inside that directory.
+
+**CLI:**
+
+```bash
+FAUXQS_DATA_DIR=./data npx fauxqs
+```
+
+**Docker:**
+
+The Docker image has `FAUXQS_DATA_DIR=/data` preset. Mount a volume and set `FAUXQS_PERSISTENCE=true` to enable persistence:
+
+```bash
+docker run -p 4566:4566 -v fauxqs-data:/data -e FAUXQS_PERSISTENCE=true kibertoad/fauxqs
+```
+
+Without `FAUXQS_PERSISTENCE=true`, the server runs in-memory even if a volume is mounted. If no volume is mounted at `/data`, persistence is silently disabled regardless of the env var (no unnecessary writes to ephemeral container storage).
+
+**Programmatic:**
+
+```typescript
+const server = await startFauxqs({ port: 4566, dataDir: "./data" });
+```
+
+All mutations are written through to SQLite immediately (no batching or delayed flush). On restart with the same `dataDir`, the server restores all state including:
+
+- SQS queues with attributes, tags, and messages (ready, delayed, and inflight with their visibility deadlines)
+- FIFO sequence counters (no duplicate sequence numbers after restart)
+- SNS topics with attributes and tags, subscriptions with attributes (fan-out works immediately after restart)
+- S3 buckets (including directory bucket type), objects with metadata, and in-progress multipart uploads
+
+`reset()` and `purgeAll()` also write through to the database — `reset()` clears messages and objects, `purgeAll()` clears everything.
+
+### Configurable queue URL host
+
+Queue URLs use the AWS-style `sqs.<region>.<host>` format. The `host` defaults to `localhost`, producing URLs like `http://sqs.us-east-1.localhost:4566/000000000000/myQueue`.
+
+To override the host (e.g., for a custom domain):
+
+```typescript
+import { startFauxqs } from "fauxqs";
+
+const server = await startFauxqs({ port: 4566, host: "myhost.local" });
+// Queue URLs: http://sqs.us-east-1.myhost.local:4566/000000000000/myQueue
+```
+
+This also works with `buildApp`:
+
+```typescript
+import { buildApp } from "fauxqs";
+
+const app = buildApp({ host: "myhost.local" });
+```
+
+The configured host ensures queue URLs are consistent across all creation paths (init config, programmatic API, and SDK requests), regardless of the request's `Host` header.
+
+### Region
+
+Region is part of an entity's identity — a queue named `my-queue` in `us-east-1` is a completely different entity from `my-queue` in `eu-west-1`, just like in real AWS.
+
+The region used in ARNs and queue URLs is automatically detected from the SDK client's `Authorization` header (AWS SigV4 credential scope). If your SDK client is configured with `region: "eu-west-1"`, all entities created or looked up through that client will use `eu-west-1` in their ARNs and URLs.
+
+```typescript
+const sqsEU = new SQSClient({ region: "eu-west-1", endpoint: "http://localhost:4566", ... });
+const sqsUS = new SQSClient({ region: "us-east-1", endpoint: "http://localhost:4566", ... });
+
+// These are two independent queues with different ARNs
+await sqsEU.send(new CreateQueueCommand({ QueueName: "orders" }));
+await sqsUS.send(new CreateQueueCommand({ QueueName: "orders" }));
+```
+
+If the region cannot be resolved from request headers (e.g., requests without AWS SigV4 signing), the `defaultRegion` option is used as a fallback (defaults to `"us-east-1"`):
+
+```typescript
+const server = await startFauxqs({ defaultRegion: "eu-west-1" });
+```
+
+Resources created via init config or programmatic API use the `defaultRegion` unless overridden with an explicit `region` field:
+
+```json
+{
+  "queues": [
+    { "name": "us-queue" },
+    { "name": "eu-queue", "region": "eu-west-1" }
+  ]
+}
+```
+
+## Supported API Actions
+
+### SQS
+
+| Action | Supported |
+|--------|-----------|
+| CreateQueue | Yes |
+| DeleteQueue | Yes |
+| GetQueueUrl | Yes |
+| ListQueues | Yes |
+| GetQueueAttributes | Yes |
+| SetQueueAttributes | Yes |
+| PurgeQueue | Yes |
+| SendMessage | Yes |
+| SendMessageBatch | Yes |
+| ReceiveMessage | Yes |
+| DeleteMessage | Yes |
+| DeleteMessageBatch | Yes |
+| ChangeMessageVisibility | Yes |
+| ChangeMessageVisibilityBatch | Yes |
+| TagQueue | Yes |
+| UntagQueue | Yes |
+| ListQueueTags | Yes |
+| AddPermission | No |
+| RemovePermission | No |
+| ListDeadLetterSourceQueues | No |
+| StartMessageMoveTask | No |
+| CancelMessageMoveTask | No |
+| ListMessageMoveTasks | No |
+
+### SNS
+
+| Action | Supported |
+|--------|-----------|
+| CreateTopic | Yes |
+| DeleteTopic | Yes |
+| ListTopics | Yes |
+| GetTopicAttributes | Yes |
+| SetTopicAttributes | Yes |
+| Subscribe | Yes |
+| Unsubscribe | Yes |
+| ConfirmSubscription | Yes |
+| ListSubscriptions | Yes |
+| ListSubscriptionsByTopic | Yes |
+| GetSubscriptionAttributes | Yes |
+| SetSubscriptionAttributes | Yes |
+| Publish | Yes |
+| PublishBatch | Yes |
+| TagResource | Yes |
+| UntagResource | Yes |
+| ListTagsForResource | Yes |
+| AddPermission | No |
+| RemovePermission | No |
+| GetDataProtectionPolicy | No |
+| PutDataProtectionPolicy | No |
+
+Platform application, SMS, and phone number actions are not supported.
+
+### S3
+
+| Action | Supported |
+|--------|-----------|
+| CreateBucket | Yes |
+| HeadBucket | Yes |
+| DeleteBucket | Yes |
+| ListBuckets | Yes |
+| ListObjects | Yes |
+| ListObjectsV2 | Yes |
+| PutObject | Yes |
+| GetObject | Yes |
+| HeadObject | Yes |
+| DeleteObject | Yes |
+| DeleteObjects | Yes |
+| CopyObject | Yes |
+| CreateMultipartUpload | Yes |
+| UploadPart | Yes |
+| UploadPartCopy | Yes |
+| CompleteMultipartUpload | Yes |
+| AbortMultipartUpload | Yes |
+| GetObjectAttributes | Yes |
+| GetBucketLocation | No |
+| ListObjectVersions | No |
+| SelectObjectContent | No |
+| RestoreObject | No |
+| RenameObject | Yes |
+| ListMultipartUploads | No |
+| ListParts | No |
+
+Bucket configuration (CORS, lifecycle, encryption, replication, logging, website, notifications, policy), ACLs, versioning, tagging, object lock, and public access block actions are not supported.
+
+### STS
+
+| Action | Supported |
+|--------|-----------|
+| GetCallerIdentity | Yes |
+| AssumeRole | No |
+| GetSessionToken | No |
+| GetFederationToken | No |
+
+Returns a mock identity with account `000000000000` and ARN `arn:aws:iam::000000000000:root`. This allows tools like Terraform and the AWS CLI that call `sts:GetCallerIdentity` on startup to work without errors. Other STS actions are not supported.
+
+## SQS Features
+
+- **Message attributes** with MD5 checksums matching the AWS algorithm
+- **Visibility timeout** — messages become invisible after receive and reappear after timeout
+- **Delay queues** — per-queue default delay and per-message delay overrides
+- **Long polling** — `WaitTimeSeconds` on ReceiveMessage blocks until messages arrive or timeout
+- **Dead letter queues** — messages exceeding `maxReceiveCount` are moved to the configured DLQ
+- **Batch operations** — SendMessageBatch, DeleteMessageBatch, ChangeMessageVisibilityBatch with entry ID validation (`InvalidBatchEntryId`) and total batch size validation (`BatchRequestTooLong`)
+- **Queue attribute range validation** — validates `VisibilityTimeout`, `DelaySeconds`, `ReceiveMessageWaitTimeSeconds`, `MaximumMessageSize`, and `MessageRetentionPeriod` on both CreateQueue and SetQueueAttributes
+- **Message size validation** — rejects messages exceeding 1 MiB (1,048,576 bytes)
+- **Unicode character validation** — rejects messages with characters outside the AWS-allowed set
+- **KMS attributes** — `KmsMasterKeyId` and `KmsDataKeyReusePeriodSeconds` are accepted and stored (no actual encryption)
+- **FIFO queues** — `.fifo` suffix enforcement, `MessageGroupId` ordering, per-group locking (one inflight message per group), `MessageDeduplicationId`, content-based deduplication, sequence numbers, and FIFO-aware DLQ support
+- **Queue tags**
+
+## SNS Features
+
+- **SNS-to-SQS fan-out** — publish to a topic and messages are delivered to all confirmed SQS subscriptions
+- **Filter policies** — both `MessageAttributes` and `MessageBody` scope, supporting exact match, prefix, suffix, anything-but (including anything-but with suffix), numeric ranges, exists, null conditions, and `$or` top-level grouping. MessageBody scope supports nested key matching
+- **Raw message delivery** — configurable per subscription
+- **Message size validation** — rejects messages exceeding 256 KB (262,144 bytes)
+- **Topic idempotency with conflict detection** — `CreateTopic` returns the existing topic when called with the same name, attributes, and tags, but throws when attributes or tags differ
+- **Subscription idempotency with conflict detection** — `Subscribe` returns the existing subscription when the same (topic, protocol, endpoint) combination is used with matching attributes, but throws when attributes differ
+- **Subscription attribute validation** — `SetSubscriptionAttributes` validates attribute names and rejects unknown or read-only attributes
+- **Topic and subscription tags**
+- **FIFO topics** — `.fifo` suffix enforcement, `MessageGroupId` and `MessageDeduplicationId` passthrough to SQS subscriptions, content-based deduplication
+- **Batch publish**
+
+## S3 Features
+
+- **Bucket management** — CreateBucket (idempotent), DeleteBucket (rejects non-empty), HeadBucket, ListBuckets, ListObjects (V1 and V2)
+- **Object operations** — PutObject, GetObject, DeleteObject, HeadObject, CopyObject with ETag, Content-Type, and Last-Modified headers
+- **Multipart uploads** — CreateMultipartUpload, UploadPart, UploadPartCopy, CompleteMultipartUpload, AbortMultipartUpload with correct multipart ETag calculation (`MD5-of-part-digests-partCount`), metadata preservation, and part overwrite support
+- **ListObjects V2** — prefix filtering, delimiter-based virtual directories, MaxKeys, continuation tokens, StartAfter
+- **CopyObject** — same-bucket and cross-bucket copy via `x-amz-copy-source` header, with metadata preservation
+- **User metadata** — `x-amz-meta-*` headers are stored and returned on GetObject and HeadObject
+- **Bulk delete** — DeleteObjects for batch key deletion with proper XML entity handling
+- **Keys with slashes** — full support for slash-delimited keys (e.g., `path/to/file.txt`)
+- **Stream uploads** — handles AWS chunked transfer encoding (`Content-Encoding: aws-chunked`) for stream bodies, including trailing header parsing for checksums
+- **Checksums** — CRC32, SHA1, and SHA256 checksums are stored on upload (PutObject, UploadPart) and returned on download (GetObject, HeadObject with `x-amz-checksum-mode: ENABLED`). Multipart uploads compute composite checksums. GetObjectAttributes supports the `Checksum` attribute. CRC32C and CRC64NVME are silently ignored. Checksums are stored and returned as-is — no body validation is performed.
+- **GetObjectAttributes** — selective metadata retrieval via `x-amz-object-attributes` header: ETag, StorageClass, ObjectSize, ObjectParts (with pagination), and Checksum (including per-part checksums for multipart objects)
+- **RenameObject** — atomic rename within directory buckets (`PUT /:bucket/:key?renameObject`). Preserves all metadata, ETag, timestamps, and checksums. Rejects general-purpose buckets. Default no-overwrite (412 if destination exists unless `If-Match` is provided). Supports source and destination conditional headers.
+- **Directory buckets** — `CreateBucket` accepts `<Type>Directory</Type>` in the body. Programmatic API: `server.createBucket("name", { type: "directory" })`. Init config supports `{ name, type }` objects in the `buckets` array.
+- **Path-style and virtual-hosted-style** — both S3 URL styles are supported (see below)
+
+### S3 URL styles
+
+The AWS SDK sends S3 requests using virtual-hosted-style URLs by default (e.g., `my-bucket.s3.localhost:4566`). This requires `*.localhost` to resolve to `127.0.0.1`. fauxqs supports several approaches.
+
+#### Option 1: `fauxqs.dev` wildcard DNS (recommended for Docker image)
+
+Works out of the box when running the [official Docker image](#running-with-docker) — nothing to configure. The `fauxqs.dev` domain provides wildcard DNS — `*.localhost.fauxqs.dev` resolves to `127.0.0.1` via a public DNS entry. Just use `s3.localhost.fauxqs.dev` as your endpoint. This replicates the approach [pioneered by LocalStack](https://docs.localstack.cloud/aws/services/s3/) with `localhost.localstack.cloud`: a public DNS record maps all subdomains to localhost, so virtual-hosted-style requests work without `/etc/hosts` changes, custom request handlers, or `forcePathStyle`. Works from any language, `fetch()`, or CLI tool.
+
+```typescript
+import { S3Client } from "@aws-sdk/client-s3";
+
+const s3 = new S3Client({
+  endpoint: "http://s3.localhost.fauxqs.dev:4566",
+  region: "us-east-1",
+  credentials: { accessKeyId: "test", secretAccessKey: "test" },
+});
+```
+
+You can also use raw HTTP requests:
+
+```bash
+# Upload
+curl -X PUT --data-binary @file.txt http://my-bucket.s3.localhost.fauxqs.dev:4566/file.txt
+
+# Download
+curl http://my-bucket.s3.localhost.fauxqs.dev:4566/file.txt
+```
+
+This is the recommended approach for host-to-Docker setups. If you are using fauxqs as an [embedded library](#programmatic-usage) in Node.js tests, prefer Option 2 (`interceptLocalhostDns`) instead — it patches DNS globally so all clients work without modification, and requires no external DNS.
+
+For **container-to-container** S3 virtual-hosted-style in docker-compose, use the [built-in DNS server](#container-to-container-s3-virtual-hosted-style) instead — it resolves `*.s3.fauxqs` to the fauxqs container IP so other containers can use virtual-hosted-style S3 without `forcePathStyle`.
+
+#### Option 2: `interceptLocalhostDns()` (recommended for embedded library)
+
+Patches Node.js `dns.lookup` so that any hostname ending in `.localhost` resolves to `127.0.0.1`. No client changes needed.
+
+```typescript
+import { interceptLocalhostDns } from "fauxqs";
+
+const restore = interceptLocalhostDns();
+
+const s3 = new S3Client({
+  endpoint: "http://s3.localhost:4566",
+  region: "us-east-1",
+  credentials: { accessKeyId: "test", secretAccessKey: "test" },
+});
+
+// When done (e.g., in afterAll):
+restore();
+```
+
+The suffix is configurable: `interceptLocalhostDns("myhost.test")` matches `*.myhost.test`.
+
+**Tradeoffs:** Affects all DNS lookups in the process. Best suited for test suites (`beforeAll` / `afterAll`).
+
+#### Option 3: `createLocalhostHandler()` (per-client)
+
+Creates an HTTP request handler that resolves all hostnames to `127.0.0.1`. Scoped to a single client instance — no side effects, no external DNS dependency.
+
+```typescript
+import { S3Client } from "@aws-sdk/client-s3";
+import { createLocalhostHandler } from "fauxqs";
+
+const s3 = new S3Client({
+  endpoint: "http://s3.localhost:4566",
+  region: "us-east-1",
+  credentials: { accessKeyId: "test", secretAccessKey: "test" },
+  requestHandler: createLocalhostHandler(),
+});
+```
+
+#### Option 4: `forcePathStyle` (simplest fallback)
+
+Forces the SDK to use path-style URLs (`http://localhost:4566/my-bucket/key`) instead of virtual-hosted-style. No DNS or handler changes needed, but affects how the SDK resolves S3 URLs at runtime.
+
+```typescript
+const s3 = new S3Client({
+  endpoint: "http://localhost:4566",
+  forcePathStyle: true,
+  // ...
+});
+```
+
+
+### Using with AWS CLI
+
+fauxqs is wire-compatible with the standard AWS CLI. Point it at the fauxqs endpoint:
+
+#### SQS
+
+```bash
+aws --endpoint-url http://localhost:4566 sqs create-queue --queue-name my-queue
+aws --endpoint-url http://localhost:4566 sqs create-queue \
+  --queue-name my-queue.fifo \
+  --attributes FifoQueue=true,ContentBasedDeduplication=true
+aws --endpoint-url http://localhost:4566 sqs send-message \
+  --queue-url http://localhost:4566/000000000000/my-queue \
+  --message-body "hello"
+```
+
+#### SNS
+
+```bash
+aws --endpoint-url http://localhost:4566 sns create-topic --name my-topic
+aws --endpoint-url http://localhost:4566 sns subscribe \
+  --topic-arn arn:aws:sns:us-east-1:000000000000:my-topic \
+  --protocol sqs \
+  --notification-endpoint arn:aws:sqs:us-east-1:000000000000:my-queue
+```
+
+#### S3
+
+```bash
+aws --endpoint-url http://localhost:4566 s3 mb s3://my-bucket
+aws --endpoint-url http://localhost:4566 s3 cp file.txt s3://my-bucket/file.txt
+```
+
+If the AWS CLI uses virtual-hosted-style S3 URLs by default, configure path-style:
+
+```bash
+aws configure set default.s3.addressing_style path
+```
+
+## Testing Strategies
+
+fauxqs supports two deployment modes that complement each other for a complete testing workflow:
+
+| Mode | Best for | Startup | Assertions |
+|------|----------|---------|------------|
+| **Library** (embedded) | Unit tests, integration tests, CI | Milliseconds, in-process | Full programmatic API: `sendMessage`, `publish`, `spy`, `inspectQueue`, `reset`, `purgeAll` |
+| **Docker** (standalone) | Local development, acceptance tests, dev environments | Seconds, real HTTP | Init config, HTTP inspection endpoints |
+
+### Library mode for tests
+
+Embed fauxqs directly in your test suite. Each test file gets its own server instance on a random port — no Docker dependency, no shared state, no port conflicts:
+
+```typescript
+// test/setup.ts
+import { startFauxqs, type FauxqsServer } from "fauxqs";
+
+export async function createTestServer(): Promise<FauxqsServer> {
+  const server = await startFauxqs({ port: 0, logger: false, messageSpies: true });
+
+  // Pre-create resources via the programmatic API (no SDK roundtrips)
+  server.createQueue("my-queue");
+  server.createBucket("my-bucket");
+
+  return server;
+}
+```
+
+```typescript
+// test/app.test.ts
+import { describe, it, expect, beforeAll, afterAll, beforeEach } from "vitest";
+
+let server: FauxqsServer;
+
+beforeAll(async () => { server = await createTestServer(); });
+afterAll(async () => { await server.stop(); });
+beforeEach(() => { server.spy.clear(); });
+
+it("tracks uploads via the spy", async () => {
+  // ... trigger your app logic that uploads to S3 ...
+
+  const event = await server.spy.waitForMessage(
+    { service: "s3", bucket: "my-bucket", key: "file.txt", status: "uploaded" },
+    undefined,
+    2000, // timeout — prevents tests from hanging
+  );
+  expect(event.status).toBe("uploaded");
+});
+```
+
+Library mode gives you deterministic assertions via the [message spy](#message-spy), non-destructive state inspection via [`inspectQueue()`](#queue-inspection), and instant state reset with `reset()` / `purgeAll()` — none of which are available from outside the process.
+
+### Docker mode for local development
+
+Use `docker-compose.yml` with an init config to give your team a consistent local environment:
+
+```yaml
+# docker-compose.yml
+services:
+  fauxqs:
+    image: kibertoad/fauxqs:latest
+    ports: ["4566:4566"]
+    environment:
+      - FAUXQS_INIT=/app/init.json
+    volumes:
+      - ./fauxqs-init.json:/app/init.json
+      - fauxqs-data:/data
+
+  app:
+    build: .
+    depends_on:
+      fauxqs:
+        condition: service_healthy
+    environment:
+      - AWS_ENDPOINT=http://fauxqs:4566
+
+volumes:
+  fauxqs-data:
+```
+
+Docker mode validates your real deployment topology — networking, DNS, container-to-container communication — and is language-agnostic (any AWS SDK can connect). The `fauxqs-data` volume persists state across restarts — queues, messages, and objects survive `docker compose down` / `up` cycles.
+
+### Recommended combination
+
+Use both modes together. Library mode runs in CI on every commit (fast, no Docker required). Docker mode runs locally via `docker compose up` and optionally in a separate CI stage for acceptance testing.
+
+See the [`examples/recommended/`](examples/recommended/) directory for a complete working example with a Fastify app, library-mode vitest tests, and Docker compose configuration.
+
+## Conventions
+
+- Account ID: `000000000000`
+- Region: auto-detected from SDK `Authorization` header; falls back to `defaultRegion` (defaults to `us-east-1`). Region is part of entity identity — same-name entities in different regions are independent.
+- Queue URL format: `http://sqs.{region}.{host}:{port}/000000000000/{queueName}` (host defaults to `localhost`)
+- Queue ARN format: `arn:aws:sqs:{region}:000000000000:{queueName}`
+- Topic ARN format: `arn:aws:sns:{region}:000000000000:{topicName}`
+
+## Limitations
+
+fauxqs is designed for development and testing. It does not support:
+
+- Non-SQS SNS delivery protocols (HTTP/S, Lambda, email, SMS)
+- Authentication or authorization
+- Cross-account operations
+
+## Examples
+
+The [`examples/`](examples/) directory contains runnable TypeScript examples covering fauxqs-specific features beyond standard AWS SDK usage:
+
+| Example | Description |
+|---------|-------------|
+| [`alternatives/programmatic/programmatic-api.ts`](examples/alternatives/programmatic/programmatic-api.ts) | Server lifecycle, resource creation, SDK usage, `inspectQueue()`, `reset()`, `purgeAll()`, `setup()` |
+| [`alternatives/programmatic/message-spy.ts`](examples/alternatives/programmatic/message-spy.ts) | `MessageSpyReader` — all spy methods, partial/predicate filters, discriminated union narrowing, DLQ tracking |
+| [`alternatives/programmatic/init-config.ts`](examples/alternatives/programmatic/init-config.ts) | File-based and inline init config, DLQ chains, `setup()` idempotency, purge + re-apply pattern |
+| [`alternatives/programmatic/queue-inspection.ts`](examples/alternatives/programmatic/queue-inspection.ts) | Programmatic `inspectQueue()` and HTTP `/_fauxqs/queues` endpoints |
+| [`alternatives/docker/standalone/`](examples/alternatives/docker/standalone/standalone-container.ts) | Connecting to a fauxqs Docker container from the host |
+| [`alternatives/docker/container-to-container/`](examples/alternatives/docker/container-to-container/) | Container-to-container communication via docker-compose |
+| [`recommended/`](examples/recommended/) | Dual-mode testing: library mode (vitest + spy) for CI, Docker for local dev |
+
+All examples are type-checked in CI to prevent staleness.
+
+## Migrating from LocalStack
+
+If you're currently using LocalStack for local SNS, SQS, and/or S3 emulation, fauxqs is a drop-in replacement for those services. Both listen on port 4566 by default and accept the same AWS SDK calls, so the migration is straightforward.
+
+There are two approaches: a Docker swap (quickest) and a hybrid setup (recommended for the best integration test experience). Which one makes sense depends on whether you use S3.
+
+### SNS/SQS only
+
+If your LocalStack usage is limited to SNS and SQS, the migration is a one-line Docker image swap. No SDK client changes are needed — the endpoint URL, port, and credentials stay the same.
+
+**Docker swap:**
+
+```yaml
+# Before (LocalStack)
+services:
+  localstack:
+    image: localstack/localstack
+    ports: ["4566:4566"]
+    environment:
+      - SERVICES=sqs,sns
+
+# After (fauxqs)
+services:
+  fauxqs:
+    image: kibertoad/fauxqs:latest
+    ports: ["4566:4566"]
+```
+
+**Region:** Both LocalStack and fauxqs auto-detect the region from the SDK client's `Authorization` header, so in most setups no configuration is needed. If you have older LocalStack configs that used the now-removed `DEFAULT_REGION` env var (deprecated in v0.12.7, removed in v2.0), the equivalent in fauxqs is `FAUXQS_DEFAULT_REGION` — it serves as a fallback when the region can't be resolved from request headers. Both default to `us-east-1`.
+
+The main difference is how resources are pre-created. LocalStack uses init hooks (`/etc/localstack/init/ready.d/` shell scripts with `awslocal` CLI calls), while fauxqs uses a declarative JSON config. `awslocal` defaults to `us-east-1` unless you pass `--region`, and fauxqs init config uses `defaultRegion` (`us-east-1`) unless you set an explicit `region` per resource — so both create resources in the same region by default:
+
+```bash
+# LocalStack init script (ready.d/init.sh)
+# awslocal defaults to us-east-1; use --region to override
+awslocal sqs create-queue --queue-name orders
+awslocal sqs create-queue --queue-name orders-dlq
+awslocal sns create-topic --name events
+awslocal sns subscribe \
+  --topic-arn arn:aws:sns:us-east-1:000000000000:events \
+  --protocol sqs \
+  --notification-endpoint arn:aws:sqs:us-east-1:000000000000:orders
+```
+
+```json
+// fauxqs init.json — uses defaultRegion (us-east-1) unless "region" is set per resource
+{
+  "queues": [{ "name": "orders" }, { "name": "orders-dlq" }],
+  "topics": [{ "name": "events" }],
+  "subscriptions": [{ "topic": "events", "queue": "orders" }]
+}
+```
+
+```yaml
+# Before (LocalStack docker-compose.yml)
+services:
+  localstack:
+    image: localstack/localstack
+    ports: ["4566:4566"]
+    environment:
+      - SERVICES=sqs,sns
+    volumes:
+      - ./ready.d:/etc/localstack/init/ready.d
+
+# After (fauxqs docker-compose.yml)
+services:
+  fauxqs:
+    image: kibertoad/fauxqs:latest
+    ports: ["4566:4566"]
+    environment:
+      - FAUXQS_INIT=/app/init.json
+    volumes:
+      - ./init.json:/app/init.json
+```
+
+Update your docker-compose service name references (e.g., `http://localstack:4566` to `http://fauxqs:4566`) and you're done.
+
+### SNS/SQS/S3
+
+When S3 is involved, the Docker swap is still straightforward — the only additional consideration is S3 URL style. If you were using `forcePathStyle: true` with LocalStack, it works identically with fauxqs. If you were using LocalStack's `localhost.localstack.cloud` wildcard DNS for virtual-hosted-style, switch to `localhost.fauxqs.dev`:
+
+```typescript
+// Before (LocalStack)
+const s3 = new S3Client({
+  endpoint: "http://s3.localhost.localstack.cloud:4566",
+  // ...
+});
+
+// After (fauxqs)
+const s3 = new S3Client({
+  endpoint: "http://s3.localhost.fauxqs.dev:4566",
+  // ...
+});
+```
+
+For container-to-container S3 in docker-compose, fauxqs includes a built-in dnsmasq that resolves `*.s3.fauxqs` to the container IP — see [Container-to-container S3 virtual-hosted-style](#container-to-container-s3-virtual-hosted-style).
+
+The init config for S3 is the same declarative JSON, with a `buckets` array:
+
+```json
+{
+  "queues": [{ "name": "orders" }],
+  "topics": [{ "name": "events" }],
+  "subscriptions": [{ "topic": "events", "queue": "orders" }],
+  "buckets": ["uploads", "exports"]
+}
+```
+
+### Going hybrid (recommended)
+
+The Docker swap gets you running quickly, but the real win comes from going hybrid: use fauxqs as an **embedded library** in your test suite and keep Docker for local development.
+
+With LocalStack, integration tests typically look like this:
+
+1. Start LocalStack container (docker-compose or testcontainers) — takes seconds
+2. Create resources via `awslocal` or SDK calls — more seconds
+3. Run your test logic
+4. Assert by polling SQS queues, checking S3 objects, etc.
+5. Clean up resources between tests — often fragile or skipped
+
+With fauxqs in library mode:
+
+1. `startFauxqs({ port: 0 })` — starts in milliseconds, in-process
+2. `server.setup({ queues: [...], topics: [...] })` — instant, no network calls
+3. Run your test logic
+4. Assert with `server.spy.waitForMessage()` — no polling, no race conditions
+5. `server.reset()` between tests — clears messages, keeps resources
+
+**What you gain:**
+
+| Concern | LocalStack Docker | fauxqs library mode |
+|---------|-------------------|---------------------|
+| Test startup | Seconds (container boot + resource creation) | Milliseconds (in-process) |
+| CI dependency | Docker required | npm only |
+| Asserting message delivery | Poll SQS queue, hope timing is right | `spy.waitForMessage()` — resolves immediately or waits |
+| Asserting message *not* delivered | `sleep()` + check | `spy.expectNoMessage()` — deterministic negative assertion |
+| Filter policy testing | Receive from queue, check absence manually | `expectNoMessage()` on filtered-out queues |
+| DLQ verification | Receive from DLQ queue via SDK | `spy.waitForMessage({ status: "dlq" })` + `inspectQueue()` |
+| Queue state inspection | `GetQueueAttributes` (counts only) | `inspectQueue()` — see every message, grouped by state |
+| State reset between tests | Restart container or re-create resources | `server.reset()` — instant, preserves resource definitions |
+| Seeding test data | SDK calls through network stack | `server.sendMessage()` / `server.publish()` — direct, no network |
+| S3 event tracking | Check bucket contents via SDK | `spy.waitForMessage({ service: "s3", status: "uploaded" })` |
+| Parallel test files | Port conflicts or shared state | Each file gets its own server on port 0 |
+
+**Migration path:**
+
+1. Install fauxqs as a dev dependency: `npm install -D fauxqs`
+2. Create a test helper:
+
+```typescript
+// test/setup.ts
+import { startFauxqs, type FauxqsServer } from "fauxqs";
+
+export async function createTestServer(): Promise<FauxqsServer> {
+  const server = await startFauxqs({ port: 0, logger: false, messageSpies: true });
+
+  server.setup({
+    queues: [{ name: "orders" }, { name: "orders-dlq" }],
+    topics: [{ name: "events" }],
+    subscriptions: [{ topic: "events", queue: "orders" }],
+    buckets: ["uploads"],
+  });
+
+  return server;
+}
+```
+
+3. Replace your LocalStack container setup with the test helper:
+
+```typescript
+// Before: LocalStack via testcontainers or docker-compose
+let endpoint: string;
+beforeAll(async () => {
+  // start container, wait for health, create resources via SDK...
+  endpoint = "http://localhost:4566";
+}, 30_000);
+
+// After: fauxqs library
+let server: FauxqsServer;
+beforeAll(async () => {
+  server = await createTestServer();
+});
+afterAll(async () => { await server.stop(); });
+beforeEach(() => { server.reset(); });
+```
+
+4. Replace polling-based assertions with spy-based ones:
+
+```typescript
+// Before: poll and hope
+await sqsClient.send(new SendMessageCommand({ QueueUrl: queueUrl, MessageBody: "test" }));
+const result = await sqsClient.send(new ReceiveMessageCommand({ QueueUrl: queueUrl, WaitTimeSeconds: 5 }));
+expect(result.Messages?.[0]?.Body).toBe("test");
+
+// After: spy knows immediately
+await sqsClient.send(new SendMessageCommand({ QueueUrl: queueUrl, MessageBody: "test" }));
+const msg = await server.spy.waitForMessage(
+  { service: "sqs", queueName: "orders", status: "published" },
+  undefined,
+  2000,
+);
+expect(msg.body).toBe("test");
+```
+
+5. Keep your `docker-compose.yml` with the fauxqs image for local development — `docker compose up` gives your team a running environment without Node.js installed.
+
+This hybrid setup gives you fast, deterministic tests in CI (no Docker required) and a realistic Docker environment for local development. See the [`examples/recommended/`](examples/recommended/) directory for a complete working example.
+
+## Benchmarks
+
+SQS throughput benchmarks are available in the [`benchmarks/`](benchmarks/) directory, comparing fauxqs across different deployment modes (in-process library, official Docker image, lightweight Docker container) and against LocalStack. See [`benchmarks/BENCHMARKING.md`](benchmarks/BENCHMARKING.md) for setup descriptions, instructions, and how to interpret results.
+
+## License
+
+MIT