Hi guys!
<h1>🌧️ mizzle</h1>
Wanted to share this package I made.
It’s the ORM Dynamo needed, it was born from the need of having the simplicity of Drizzle with Dynamo.
So I thought to myself, why not make an ORM and make it feels like drizzle.
With that in mind mizzle was born.
Mizzle provides a type-safe, fluent API for interacting with DynamoDB, supporting relational queries, batch operations, and transactions.
<h2>Key Features</h2>
<ul>
<li>Type-Safe Schema Definition: Define your tables and entities with strict TypeScript types.</li>
<li>Fluent Query Builder: precise API for <code>insert</code>, <code>select</code>, <code>update</code>, and <code>delete</code> operations.</li>
<li>Relational Queries: Query related entities with <code>db.query</code>.</li>
<li>Batch Operations: <code>batchGet</code> and <code>batchWrite</code> support.</li>
<li>Transactions: Atomic operations using <code>db.transaction</code>.</li>
<li>Automatic Type Inference: <code>InferSelectModel</code> and <code>InferInsertModel</code> utilities.</li>
</ul>
<blockquote>
Skip this with the <a href="https://mizzle-docs.vercel.app" target="_blank" rel="noopener nofollow">way less boring docs</a>
</blockquote>
<h2>🚀 Installation</h2>
<pre><code class="language-bash">npm install @aurios/mizzle
# or
bun add @aurios/mizzle
</code></pre>
<h2>Get Started</h2>
<h3>Defining the Table</h3>
In DynamoDB, you first need a physical table. Mizzle separates the definition of the physical table structure (PK, SK, Indexes) from the logical entities that live within it and with this making your database well organized and easier to reason about.
<pre><code class="language-ts">import { dynamoTable, string } from &quot;@aurios/mizzle&quot;;

// This matches your actual DynamoDB table configuration
export const myTable = dynamoTable(&quot;JediOrder&quot;, {
 pk: string(&quot;pk&quot;), // The partition key attribute name
 sk: string(&quot;sk&quot;), // The sort key attribute name (optional)
});
</code></pre>
<h3>Defining the Entity</h3>
An Entity represents your data model (e.g., an user, an item on an user). You map the Entity to a Physical Table and define how its keys are generated, so every entity looks kinda like an separated table.
<pre><code class="language-ts">import { dynamoEntity, string, uuid, number, enum, date, prefixKey, staticKey } from &quot;@aurios/mizzle&quot;;

export const jedi = dynamoEntity(
 myTable,
 &quot;Jedi&quot;,
 {
 id: uuid(), // Automatically generates a v7 UUID
 name: string(),
 homeworld: string()
 },
 (cols) =&gt; ({
 // PK will look like &quot;JEDI#&lt;uuid&gt;&quot;
 pk: prefixKey(&quot;JEDI#&quot;, cols.id),
 // SK will be a static string &quot;PROFILE&quot;
 sk: staticKey(&quot;PROFILE&quot;),
 }),
);

export const jediRank = dynamoEntity(
 myTable,
 &#x27;JediRank&#x27;,
 {
 jediId: uuid(),
 position: string().default(&#x27;initiate&#x27;),
 joinedCouncilDate: string(),
 },
 (cols) =&gt; ({
 // Same as the jedi so they can be related
 pk: prefixKey(&#x27;JEDI#&#x27;, cols.jediId),
 // RANK#initiate
 sk: prefixKey(&#x27;RANK#&#x27;, cols.position),
 })
)
</code></pre>
<blockquote>
The UUID V7 was chosen for better sorting, since the values are time-sortable with 1 millisecond precision.
</blockquote>
<h3>Define the Relations</h3>
Relationships in Mizzle are established using the defineRelations function. This step creates a logical map of how your entities interact, enabling you to perform powerful relational queries—such as fetching a Jedi along with their entire rank history—in a single operation.
<pre><code class="language-ts">import { defineRelations } from &quot;@aurios/mizzle&quot;;
import * as schema from &quot;./schema&quot;;

export const relations = defineRelations(schema, (r) =&gt; ({
 jedi: {
 // A Jedi could&#x27;ve a lot of ranks throughout his year
 ranks: r.many.jediRank({
 fields: [r.jedi.id],
 references: [r.jediRank.jediId],
 }),
 },
 jediRank: {
 // Every registry points to one Jedi only
 member: r.one.jedi({
 fields: [r.jediRank.jediId],
 references: [r.jedi.id],
 }),
 },
}));
</code></pre>
<h3>Initialization</h3>
Initialize the mizzle client by passing it an instance of the standard AWS DynamoDBClient and the relations we just defined.
<pre><code class="language-ts">import { DynamoDBClient } from &quot;@aws-sdk/client-dynamodb&quot;;
import { mizzle } from &quot;@aurios/mizzle&quot;;
import { relations } from &quot;./relations&quot;;

const client = new DynamoDBClient({ region: &quot;us-east-1&quot; });
export const db = mizzle({ client, relations });
</code></pre>
<h3>Query</h3>
Now you can use the fluent API to interact with your data in the best way possible.
<h4>Insert Data</h4>
<pre><code class="language-typescript">// api/jedi/new.ts
import { jedi } from &quot;$lib/schema.ts&quot;;

const newJedi = await db
 .insert(jedi)
 .values({
 name: &quot;Luke Skywalker&quot;,
 homeworld: &quot;Tatooine&quot;,
 })
 .returning();

console.log(newJedi.id); // The auto-generated UUID
</code></pre>
<h4>Select Data</h4>
Mizzle intelligently routes your request to <code>GetItem</code>, <code>Query</code>, or <code>Scan</code> based on the filters you provide.
<pre><code class="language-typescript">// /api/jedi/get.ts
import { jedi } from &quot;$lib/schema.ts&quot;;
import { eq } from &quot;@aurios/mizzle&quot;;

// This will use GetItem because both PK and SK are fully resolved
const user = await db.select().from(jedi).where(eq(jedi.id, &quot;some-uuid&quot;)).execute();
</code></pre>
<h2>Mizzling</h2>
This will work if you already has a DynamoDB table with data in it. If you want to create a new table with the schema you defined, you can use the <a href="https://www.npmjs.com/package/@aurios/mizzling" target="_blank" rel="noopener nofollow"><code>mizzling</code></a> CLI.
I truly hope this helps someone else as it helps me with my other projects ❤️

Lucas A. Ouverney

Mizzle is a new type-safe ORM for DynamoDB inspired by Drizzle's developer experience. It provides a fluent API for defining schemas, entities, and relationships with TypeScript type inference. Key features include relational queries, batch operations, transactions, and automatic routing between GetItem, Query, and Scan operations. The library separates physical table definitions from logical entities, uses UUID v7 for time-sortable keys, and includes a CLI tool for table creation.

Mizzle - A drizzle like ORM for DynamoDB