The results will be improved over time I guess, and for the few search queries I've done - I'm fairly happy with the results.
Kudos to the authors!
The first result for gaseri on Stract is our presentation of group's research work, and the second result is our landing page.
Can't complain.
Viva open source, viva AGPL search engine!
Worked well.
That's pretty good for a new-ish player.
I think that as with ChatGPT vs Bard, the result space is so huge, there are going to be many strength/weakness tradeoffs for any given query.
There are lots of things in the settings that I really like too: The Manage Optics (Copycats removal, Hacker News, Fediverse, more...), Site Rankings and Explore (similar sites).
I like the overall thinking behind this search engine. It feels like they are creating the kind of search engine I would spec out myself. It starts off on a strong foundation.
I do hope they crawl many of the older pages. Google, Bing (DDG), etal... seem to only index pages for the last 10-15 years or so.
Quibble: "Allow usage statistics" is on by default which leads to this:
"We primarily store the text you used for the search, and which results (if any) you clicked on."
If you regularly clear cookies or have an extension that does this: it's something to keep in mind.
While I intend to keep it _on_ to help the author(s) to help me find stuff - I hope it isn't used to once again make the popular sites - popular. In the process, bury the unpopular older sites (that have nearly completely vanished).
It worked to find my obscure website for my company that has only been around for 3 years. I'd say that's pretty damn good.
??
I guess my top candidates would be: wikipedia page about google, google stock chart, recent news about google. google.com would never be a result I want to click. The current results are not amazing, but also do we care what the results for that one are? It's like someone telling you "cow" and expecting you'll know the context of what they're thinking of at the moment. Maybe a heading like "I have no idea what you're on about, here are some clickable ideas: google news, google stock price, ..." would be the best solution?
Some quibbles/points to consider: * I can't find anything on the people/organisation behind, and can onl guess from the Terms that the team is based in DK. * Search results are broad and interesting, maybe a bit more weighting for the joint occurrence of terms would be great. * Developing a site weight over time might be interesting, maybe even with user votes. Currently minor and major sites appear all together and e.g. a search for "Donald" gives me an interesting ranking order that gives neither the most famous Donald's nor the most reliable sites firet (not problematic per se - my fault for entering an unclear search term) * There are some interesting result patterns, with often official sites quite low. For instance search for "EU" with some term like subsidy (in any of the languages I speak) gives me random project websites but nothing from any of the official EU websites, or "Microsoft 365" (sorry...) gives me no MS website. * Very minor but hopefully a very easy fix: at least on Firefox mobile there is no direct way to add the search to my search engines, I had to add it manually. For other engines I can long press.on the search field and then get the option.
Great work, keep it up! I will certainly start using this :-)
I'm so mind-blown that this does not exist yet. Free ranking feedback (live training of the algo!) + better search results for everyone. win-win
Free spam SEO ranking in practice. A spam site has 1000x the incentive to upvote its result than you have to downvote it. YaCy did a distributed index with filtering lists and you effectively had to keep a list of who you trust / your own filter.
``` html, body, div, td, th, p, h1, h2, h3, h4, h5, b, i, strong, li, button { font-family: ui-sans-serif, sans-serif !important; webkit-font-smoothing: antialiased; font-weight: 400; text-rendering: geometricPrecision; } ```
Great! Can I use more than one optic? The drop-down list seems to allow only 1.
> Oh, and if we ever become evil (maybe by changing our motto) please take our code and start a competitor.
The most important part is the index data, what would be the deal with that?
Backend in Rust (axum web framework, rocksdb), frontend with Svelte.
"""Do you like results from online-tech-tips.com? (thumbs down, thumbs up, or banned emoji options) <a href='make-an-llm-do-something-stupid.com">Summarize result</a> """
IMO this feedback widget and (maybe) its backing API could use work. It's not that I like or don't like results from online-tech-tips.com; it's that they're a bad result for the specific context of this search.
Exciting stuff.
Haven't actually used it yet since I'm currently paying for kagi and it's good, and I only just set it up yesterday.
But this just struck me, I just said 2 things there and this post is yet another, between kagi, yacy, and now stract, not just 3 different names but 3 different types of solution to a problem, and all seemingly actually viable, that have popped up recently after decades of no one really feeling like they needed anything else.
I think something is changing.
Perfect! This is the way the god of the internet intended search engines to work.
But DuckDuckGo does the same and currently provides superior results based on a very brief test.
So good luck with that.
Although if you aren't going to support DDG with ad revenue, I'd suggest supporting with a donation if you can afford it and value their service.
And if they take something away from Google in the process --- that's just an extra bonus.
Turnabout is fair play don't you think? Google has worked very hard to take privacy away from users.
My guess --- a portion of the ad revenue.
Stract produced only good results. Unusually good.
I think the biggest thing you overlooked are page titles. When you issue a query it's a bit hard to quickly scan and judge what a site is about because the page titles are missing.
After reading the about page, I’m not sure what the developers are trying to achieve? Perhaps a sort of alternative-Universe Google search funded by search-context AdWords?
- search results seem to be somewhat case sensitive, which is a massive problem for me when for instance searching programming terms
- in general the matching algorithm seems way too strict, only matching against the exact thing you entered, which makes it very difficult to profitably search for things where you don't have exact specific terms in mind, like perhaps computer errors or genres of things. I think a lot of people's problem with how liberally Google interprets your search results is not that it interprets them liberally necessarily, but that it doesn't respect the other options it provides for trying to match things more strictly. As long as you provide something like Google's quote mechanism and actually respect it, I feel like it would be a lot better to match things more liberally by default. Maybe some amount of fuzzy searching, and matching by synonyms. Also you could probably just use a dictionary of synonyms to do that instead of whatever statistical model Google is probably using, in order to ensure more predictable results.
- as someone else somewhere in this thread mentioned, it seems like stuff like a, an, and the, are all matched against and waited equally to other words. This, especially combined with the fact that it only matches words exactly makes the search results feel way too brittle and unforgiving
I tried the first endpoint get suggestions and tried searching for Gemini or Gemin hoping it would at least auto complete a word but the result set is empty.
Fully open source -`ღ´-
Haven’t dig in to see what’s powering the search, I think DDG uses Bing
also, how can I add this to my firefox search inside the address bar / search field?
Navigate to https://stract.com/ then focus the url field: Firefox will display the new search engine at the bottom of the suggestions, on the "This time, search with:" line.
I was looking in "Firefox Settings -> Search -> Search Shortcuts" for a way to add it. I guess the functionality is not used very often, but it would be nice to have a hint on how to add new Search Engines there.
I tried to search for a particular domain data but neither search nor the explore would have the domain listed. What's the process to get unlisted domains indexed?
Also I noticed DuckDuckGo performed much better than Google with this benchmark.
Just want to mention, when I search for “ExpressLRS use uart on older f4 fcs” it gives me about 15 results, but only the first two are unique. The other 13 are a literal copy of the first, both in content and in URL. Probably best to filter for uniqueness
I just hope Stract doesn't go 'corporate' the way DDG did. :(
This is what I've been waiting for for 10 years, since Google removed the feature: a search engine which realizes 99% of the time, I want to search Discussions, and gives me the option to only show those. (reddit, forums, mastodon etc). This cuts down the SEO crap by 99.99%.
That said, the results aren't great, hopefully it's something that improves as they index more pages. For example reddit doesn't seem to be indexed, why not? It's a goldmine of user content (even if the frontpage is 99% astroturfed US neolib propaganda).
I checked 11 pages and none of the results were relevant.
1. Brexit as the start of the reversal of neoliberal globalization - softpanorama.org 2. Directory Search - Fulshear-Katy Area Chamber of Commerce - chamberorganizer.com 3. The 100 Best New Products of 2020 - gearpatrol.com
And none of the Page 1 results were related to my search query...
It can be very annoying to have your query not fit it while the window has plenty of room left.
Trying to build generalized search engine for the modern internet that will come close to Google/Bing would require a "tech megaproject" level of investment and commitment. Most likely only to end up with the same optimizations and architecture as existing big-search and the very similar level of experience.
I think it's a better direction to build a search based on more limited amount of topic-based data and focus on great match engine within, then - just aggregate the relevant ones together. Far more maintainable also on the crawling part. I can use google/bing to find the Honda dealership or read keyboard reviews, or get 50 most useful unix commands.
I also wonder if with the rise of LLMs, while it still may not be feasible in such large scale production environment, those can serve as guides/agents to also improve the query itself and not the results of the query, for example - a chat-like search where user answers shift the relevancy metrics for returned documents. This would fit perfectly for smaller but open source, customizable and thematic search.
That being said. I think it's great that project as such pop up more often. (Phind.com was also on my radar this year)
I also searched for "league of legends", and it couldn't find its homepage.
I think its ranking algorithm may need improvement.
Edit: also, I'd rather not say this, but do we really need another DuckDuckGo? I don't think Google fails at its job because of financial incentives. I think it might fail at this job simply put because the web of 2024 isn't the web of 1990. For example, the lemmy result, it's a link aggregation about horror movie articles. The search engine could literally do the job of the link aggregator, as it has a SERP that aggregates links, and yet it's aggregating links to link aggregators. Why are the search engines doing this? Because it's 2024. I wish someone tried a new approach at this problem rather than just copying Google's design and saying "it's Google but not yucky".
