We’re (a group of friends) building a search engine from scratch to compete with DuckDuckGo. It still needs a name and logo.

Here’s some pictures (results not cherrypicked): https://imgur.com/a/eVeQKWB

Unique traits:

  • Written in pure Rust backend, HTML and CSS only on frontend - no JavaScript, PHP, SQL, etc…
  • Has a custom database, schema, engine, indexer, parser, and spider
  • Extensively themeable with CSS - theme submissions welcome
  • Only two crates used - TOML and Rocket (plus Rust’s standard library)
  • Homegrown index - not based on Google, Bing, Yandex, Baidu, or anything else
  • Pages are statically generated - super fast load times
  • If an onion link is available, an “Onion” button appears to the left of the clearnet URL
  • Easy to audit - No: JavaScript, WASM, etc… requests can be audited with F12 network tab
  • Works over Tor with strictest settings (official Tor hidden service address at the bottom of this post)
  • Allows for modifiers: hacker -news +youtube removes all results containing hacker news and only includes results that contain the word “youtube”
  • Optional tracker removal from results - on by default h No censorship - results are what they are (exception: underage material)
  • No ads in results - if we do ever have ads, they’ll be purely text in the bottom right corner, away from results, no media
  • Everything runs in memory, no user queries saved.
  • Would make Richard Stallman smile :)

THIS IS A PRE-ALPHA PRODUCT, it will get much MUCH better over the coming months. The dataset in the temporary hidden service linked below does not do our algorithm justice, its there to prove our concept. Please don’t judge the technology until beta.

Onion URL (hosted on my laptop since so many people asked for the link): ht6wt7cs7nbzn53tpcnliig6zrqyfuimoght2pkuyafz5lognv4uvmqd.onion

  • sqw@lemmy.sdf.org
    link
    fedilink
    English
    arrow-up
    3
    ·
    8 months ago

    i feel that decentralized search is an extremely valuable thing to start thinking about. but the devil is in practically every one of the details.

    • sugar_in_your_tea@sh.itjust.works
      link
      fedilink
      arrow-up
      1
      ·
      8 months ago

      Yup. Even if you trust all your peers (which isn’t reasonable), there’s still a ton of practical issues that need to be resolved:

      • pagination with a different set of peers
      • moderation of CSAM and whatnot
      • outdated peers and stale data
      • how much data and where are results reduced

      It’s a really complex problem without getting p2p involved, and p2p just adds a ton of other problems.

      So I’m probably going to stick with building my Reddit clone, which I think is simpler (search doesn’t need to happen at the start).