nat

Zzut · 3 Mar

I noticed that when you have the assistant summarize a site for example, there will be around 10 requests made by at least 2 different search engine crawlers. Google being the most aggressive one.

I don't think this is a good thing, if people use Kagi's assistant in for example professional setting, this will eventually lead into leaking secrets to these search engines. I kinda assumed that Kagi had some privacy with the API's it uses from these providers, like with the LLM API's.

Toggling the web access off from assistant is a "fix" to this, but this seems like it's an issue that might need some consideration.

kzar · 7 Feb

Similar to the AI summariser tool, I'd like a tool to read articles out using a text to speech model. That would be really helpful, as I find podcasts and audiobooks much easier to digest. Bonus points if the feature has a way to provide a personal RSS feed, that can then be hooked into a podcast app, but even just a more manual page to convert articles would be useful.

I would leave tabs open in my mobile browser, for interesting articles that I haven't yet got around to reading. I'd love to be able to listen to those articles, while doing the dishes or other chores.

LLoganKaser · 21 Jan

Under Firefox, clicks on results to expand the image result in:

Too many calls to Location or History APIs within a short timeframe. k_ima.min.js:1:19232
Uncaught DOMException: The operation is insecure.
    onclick https://kagi.com/asset/v2/js/k_ima.min.js?v=f7409af16a06a7643ac6b103e8271bf557f6b553:1
    initImagePreview https://kagi.com/asset/v2/js/k_ima.min.js?v=f7409af16a06a7643ac6b103e8271bf557f6b553:1
    renderImagesPage https://kagi.com/asset/v2/js/k_ima.min.js?v=f7409af16a06a7643ac6b103e8271bf557f6b553:1
    <anonymous> https://kagi.com/asset/v2/js/k_ima.min.js?v=f7409af16a06a7643ac6b103e8271bf557f6b553:1
    scriptStreamDispatchEvent https://kagi.com/images?q=flush baseboard:1
    scriptStreamCallback https://kagi.com/images?q=flush baseboard:1
    <anonymous> https://kagi.com/images?q=flush baseboard:1015
k_ima.min.js:1

Or:

Uncaught (in promise) DOMException: Invalid encoded image data.

For roughly 80% of result images.

Basic usage of the image search (clicking on images to expand them) to work.

drguildo · 20 Jan

When clicking on certain image thumbnails in image search results, the image fails to open. I get the following in the developer console:

The strange thing is that resizing the window seems to "fix" the issue for those images, but then other images fail to open. The search terms don't seem to matter and it happens with any result set.

I'm using Firefox 134.0.1 (64-bit) on Windows 10. I also get the same behaviour on Fedora Linux.

SSyneRyder · Sep 4, 2024

I'd like to see an option to exclude specific external indexes from my search.

The benefits I see are these:

Improved Search Quality
I recently had a Kagi search result that included 20 duplicate results, all from different proxy / "unblocking" domains that had scraped the same site. A terrible search experience. I was able to block those domains, but I'm playing whack-a-mole doing so. The same search on Google, Brave, Mojeek or Bing didn't include any of those duplicate proxy results. I think I have identified which external index those proxy results came from, and I would like to exclude that index for all my future results.

Helps De-Googlers & De-Bingers
Some users choose Kagi specifically to escape Google & Microsoft, and are stunned when they learn Kagi uses Google APIs to generate their results. Being able to exclude Google as an external index would help these users.

Helps Users With Ethical Concerns
Some users have previously said they would prefer not to use the Brave index. Some would prefer not to use the Yandex index. And some want access to every index they can. The politics here aren't actually relevant - what is relevant is that users can't configure their own Kagi account to work the way they want it to. Excluding specific indexes / API calls lets users ensure their money goes only to the external sources they want to fund.

Makes Kagi More Transparent
The current Search Sources page on Kagi is vague, and doesn't directly mention that Google, Bing, Brave and Yandex are index sources. Listing all the sources would help users know what is going on. It is unlikely to give any secret recipe information to competitors that they didn't know already.

Incentivizes Building An Independent Index
I actually thought Kagi was an independent index when I signed up (I misunderstood how limited Teclis is). I thought my subscription money was going towards growing a new independent service. Brave has proven it is possible for a newcomer to build a 100% independent index.

I would envision a new "Indexes" page, under Kagi -> Settings -> Search -> Indexes. There it would list all of the external indexes that Kagi uses, with a checkbox next to each. There might also be a sentence under each that briefly explains the index (eg Includes results from Google Search, and also Google Images) and with a link to the index (eg mojeek.com) for users to research that index themselves.

SearX already has something like this, in its Preferences -> Engines page, with switches next to each search engine to enable or disable:

When the index is disabled, ideally Kagi would not issue API requests to that index service on my behalf. The purpose is to ensure I'm not unwittingly funding those companies.

When new indexes are added, I would expect them to be enabled by default. Since some users may object to this, you might also need to add an option to uncheck "Enable New Indexes By Default" on that page as well.