Self Hosted Search Stack
I have finally modified my self hosting stack to include search! Hooray!
As my takes have evolved on AI, I have slowly decided I did not want to involve generative AI in my searches; at least not from a search engine perspective. I do use AI (those are my takes). I self host some basic models. I am selective with my AI usage.
Why does this matter? Well, I found myself going down, what I would call are AI driven rabbit holes. I would be searching for a specific issue - a lot of it had to do with troubleshooting - what I would receive is a result specific to a hardware, technology, even unique software taken out of context to the basic configuration or troubleshooting prompt I would be submitting. Inherently, I would go down the hole trying to enhance the output ending up reviewing minorly relevent links that would not lead to the resolution to my problem.
The Full Stack
I want to highlight the entire stack. They all work together to produce my entire search experience.
- Wireguard: All of my devices have a VPN profile with different permissions. If they have access to the DNS, they have access to the search engine.
- PiHole + Unbound: I use a custom PiHole and Unbound container image for a self-hosted DNS solution. This is where I configure the blocked domains as well as the custom domains such as the one for my search.
- SearXNG: The search engine and aggregation software doing the search magic.
- Open WebUI: Wait, isn't that an AI tool? Yes, it is. I have found this very effective for non important searches for general interests. The two ways I have used this the most are searching for upcoming movies as well as song lyrics.
- Browsers: I am using Vivaldi and Brave, the key point to highlight here is changing the default search. I am a frequent taskbar searcher, so I put no domains in the bar all the time. They use the string to query the default browser.
How Effective Has It Been?
It has been better than expected. It is a completely different search experience. It feels like retro, solar punk. You get the old school experience with modern day technology and features. Like I stated earlier, it aggregates, you get to see which engines had similar results. You can enhance your privacy and optimize your usage by refining the sources of results. You can use different sources for different types of findings.
Would You Recommend It?
Yes: You are into self hosting. You care about privacy. You don't like AI. You like your self-hosted AI. You like to tinker or want to tinker more. You miss the 90s.
No: You don't care about the above. You benefit from modern day AI search. You are very non-technical. You have no desire to learn about this space.
The "No" category may seem harsh, it is not meant to be. People have different desires and this type of activity isn't for everyone.
Definitely a fun project and path forward for a lot of people interested in having more control over your search!
