---
title: Compress
description: Compresses responses with zstd, brotli or gzip, as the client accepts.
---
> ## Documentation Index
> Fetch the complete documentation index at: https://docs.page/luizfonseca/super/llms.txt
> Use this file to discover all available pages before exploring further.

Compresses the responses of a location with zstd, brotli or gzip. Of the algorithms the client accepts (`Accept-Encoding`), the best one is used: zstd, then brotli, then gzip, whatever order the client lists them in. `Vary: Accept-Encoding` is added, so caches keep one copy per encoding.

## Options

| Name | Description |
| --- | --- |
| `algorithms` | The algorithms the location uses, among `"zstd"`, `"br"` and `"gzip"` (default: all three) |
| `level` | The compression level of every algorithm (default: `3` for zstd, `4` for brotli, `6` for gzip, a balance of speed and size). Up to `22` for zstd, `11` for brotli and `9` for gzip. |
| `priority` | When it runs among the blocks of the location (default: `40`) |

### Usage

**config.hcl**
```hcl
host "mywebsite.com" {
  location "/" {
    upstream "localhost:3000" {}

    compress {}
  }

  location "/api" {
    upstream "localhost:4000" {}

    compress {
      algorithms = ["zstd", "gzip"]
    }
  }
}
```

### Good to know

- Responses are compressed as they are proxied, without waiting for the whole body.
- Text, JSON, JavaScript, fonts, SVG and similar types are compressed. Responses smaller than 20 bytes, images, video and archives are sent as they are.
- Responses the upstream already compressed are sent as they are.
- Responses from the [cache](/use-cases/cache) are compressed as they are sent, so the cache keeps a single copy.
- Higher levels make responses smaller but cost more CPU per response: the defaults suit responses compressed as they are sent.