> ## Documentation Index
> Fetch the complete documentation index at: https://docs.e6data.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Native Engine

> workers.runtime.duckdb

ReferenceApplies to: Unified engine deployments

`[workers.runtime.duckdb]`

## Settings

### `extension_paths`

Lists already installed Native Engine extension files to load.

* **Default:** `[]`
* **Type:** array

> **Example**
>
> `["@bundled/icu"]` selects the ICU extension shipped with the package. An entry here does not download or install an extension.

### `threads`

Sets the maximum number of execution threads the Native Engine may use before task-specific limits apply.

* **Default:** `2`
* **Type:** integer; positive

> **Example**
>
> `2` allows up to two engine execution threads. Setting `8` does not allocate eight CPUs to the worker pod.

### `memory_bytes`

Sets the Native Engine’s memory budget in bytes.

* **Default:** `1073741824`
* **Type:** integer (bytes)

> **Example**
>
> `1073741824` sets a 1 GiB engine budget. The worker still needs memory for connectors, output buffers, and other process work. Zero removes this engine cap.

### `temporary_storage_bytes`

Limits how much temporary disk storage the Native Engine may use.

* **Default:** `4294967296`
* **Type:** integer (bytes); optional

> **Example**
>
> `4294967296` sets a 4 GiB spill cap. Zero removes the cap; it does not disable disk spill.

### `output_buffer_batches`

Limits how many completed output batches wait between the Native Engine and its consumer.

* **Default:** `2`
* **Type:** integer; positive

> **Example**
>
> `2` allows two batches to wait. More batches may smooth delivery but also keep more result data in memory.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.