warehouse/blog/selfhost-search-engine.typ

#import "template.typ": conf
#show: doc => conf(
  title: [ Host your own private search engine with SearXNG ],
  doc,
)

= Introduction
#link("https://docs.searxng.org/")[SearXNG];, put in its own words, is a
'free internet metasearch engine'. Note that it describes itself as a
#emph[metasearch] engine specifically - unlike your traditional search
engine like Google or Bing, SearXNG does things a little bit
differently: It aggregrates the results produced by search services like
those aforementioned, and feeds them back to you.

Because of this key detail and a great deal of effort by those who’ve
helped shape it, SearXNG protects your privacy, and does so very well: -
Private data from requests going to the search services it aggregrates
results from is removed - It does #strong[not] forward #emph[anything]
to any third parties through search services - Private data is
#emph[also] removed from requests going to the results pages

Furthermore, SearXNG can be configured to use
#link("https://torproject.org")[Tor];.

However, the aspect of privacy isn’t the only great selling feature of
the engine; from my use of the engine so far, it’s also great
at…searching \(is that a surprise?). The fact that it’s a metasearch
engine plays a key role in this, as it provides SearXNG the ability to
pull content more efficiently and gives #emph[you] the ability to
further tailor your experience.

= Setting up SearXNG
== Installing the service
As you may have expected if you’ve used NixOS for a while, searxng is
packaged #emph[and] has a service on NixOS. This makes setting it up
just that much easier.

To get started, place somewhere in your #emph[system] config the
following:

```nix
{
    # ...
    services.searx = {
    enable = true;
      settings = {
        server = {
          port = 8888;
          bind_address = "127.0.0.1";
          secret_key = "@SEARX_SECRET_KEY@";
          base_url = "https://search.devraza.duckdns.org/"; # replace with wherever you want to host yours
        };
      };
    };
    # ...
}
```

The snippet above starts the `searx` systemd service for listening on
port `8888`, and assumes a `base_url` of
`https://search.devraza.duckdns.org`.

Now that we’ve got the actual `searx` instance running, we can now set
up a reverse proxy allowing the service to be accessed remotely
\(whether this is within your local network or across the internet is up
to you).

== Setting up a reverse proxy
=== What is a reverse proxy?
Before I get started with the technical details of setting this up, I’d
like to briefly clarify what a reverse proxy exactly is \(to my
understanding).

Let’s get the wikipedia definition of reverse proxy out of the way
first:

#quote(block: true)[
\[…\] a reverse proxy is an application that sits in front of back-end
applications and forwards client requests to those applications. \[…\]
]

However, you might be confused as to what this actually means; I’ll give
an example of the usage of reverse proxies to better explain this:

- Suppose you’ve got a few services running on a server \(for
  demonstration purposes, let’s name these `x`, `y` and `z`), each
  running on their own unique port.
- Assuming you had a domain, and wanted to access all of these services
  from their own unique sub-domains \(e.g.~`x.yourdomain.com`,
  `y.yourdomain.com` and `z.yourdomain.com`), you would have to use a
  reverse proxy.
- This reverse proxy would take in requests from clients going to
  sub-domains, and forward these requests to the appropriate port on
  your machine for the service being requested.

The concept should be clear now, if it wasn’t already.

=== Using NGINX to set up the reverse proxy
NGINX is a popular web server that supports the creation of virtual
hosts and the usage of reverse proxies. To accomodate our `searx`
instance, we append the following to our NixOS server configuration:

```nix
{
  # ...
  services.nginx = {
    enable = true;
    # any extra configuration here
    virtualHosts = {
      "search" = { # this can be anything, being an arbitrary identifier
        forceSSL = true;
        serverName = "search.yourdomain.com"; # replace this with whatever you're serving from
        # SearX proxy
        locations."/" = {
          proxyPass = "http://${toString config.services.searx.settings.server.bind_address}:${toString config.services.searx.settings.server.port}";
          proxyWebsockets = true;
          recommendedProxySettings = true;
        };
      };
    };
  };
  # ...
}
```

The expression highlighted above is used to dynamically adjust the location NGINX will forward requests to, depending on your `searx` config

After saving your changes and rebuilding your server’s system
configuration \(as usual), you should have a working #emph[private]
instance of SearXNG that you can access using the `serverName` you’ve
given it.

Set your browser to use this as your search engine using the relevant
documentation \(with Firefox this is as easy as right-clicking on the
URL after opening up the page and clicking a button). Enjoy!