Assistant error "something went wrong..." when "web access" is selected

mille

Not much to add, but happens to me, too. Across devices (Firefox on Linux, Firefox on MacOS) and with all backends (Claude, ChatGPT, Gemini, etc.). For me, this happens always, and is obviously pretty annoying because I have to F5 until I can finally see the complete answer (instead of spending my time reading the answer as it is returned, which is more efficient).

yiwei-1

mille

Could you share some thread IDs where this happens? You can just post the links here without sharing the content and we'll check the error logs. You can also DM us the threads on Discord if you prefer.

mille

yiwei-1

Sure, here you go:

https://kagi.com/assistant/9d3c3fc6-6952-4640-971d-6effe12ef52b (Deepseek Chat v3)
https://kagi.com/assistant/00cf7520-09a5-47d9-b9b8-b120d633dfd2 (Gemini Flash 2)

Nyaa

yiwei-1 https://kagi.com/assistant/bec9f369-c1da-4563-ba02-859f44f46ec7

Nyaa

yiwei-1 Could you share some thread IDs where this happens?

yiwei-1 https://kagi.com/assistant/9dddfc10-ed34-486f-a698-a550d45b2e89

Nyaa

yiwei-1 https://kagi.com/assistant/c7307c46-5ea8-4cde-998a-a6a8ef830556

mille

~~For me, the problem disappeared completely a couple of days ago. I suppose this is something you guys fixed, so thanks a bunch! 👏 🙏~~

Nevermind, the problem is back. That means problem is also intermittent for me.

mille

Unfortunately, even with the new version I still get

Nyaa

So i now get this message. While it’s not bad to show some message to the user, the content is…. wrong.
I have a very stable internet connection (my country is in the top10 of best internet connections in the world by the way), and i’m mostly not even on Wi-Fi, and have 0 issues with internet connectivity.

Kagi Assistant gives me this error daily.

Contrary to your “assumption” (not based on facts), the issue is NOT my “connection with the internet”…

yiwei-1

Nyaa

Did you happen to use a VPN by the way? We’ve seen some cases in the past that using a VPN can sometimes result in a less stable connection to our server

mille

yiwei-1 Can't answer for @Nyaa, but yes, I'm always on ProtonVPN.

Same for me, BTW: I'm on a very stable, cabled fiber connection.

Nyaa

yiwei-1 Yea, i also often use ProtonVPN. So that might be the issue. How can i help solve this? I can reproduce easily, so if you want me to run any test let me know;

Note that the connection to kagi.com seems to be stable. Also over a long period of time. This is a clean Safari profile, without any extensions

mille

Ah, I see. When on ProtonVPN using Wireguard, the TLS connection is broken after a couple of seconds:

H=kagi.com; time openssl s_client -quiet -verify_return_error -servername $H $H:443 
Connecting to 34.111.242.115
depth=2 C=US, O=Google Trust Services LLC, CN=GTS Root R1
verify return:1
depth=1 C=US, O=Google Trust Services, CN=WR3
verify return:1
depth=0 CN=kagi.com
verify return:1
80CB5B021D740000:error:0A000126:SSL routines::unexpected eof while reading:ssl/record/rec_layer_s3.c:696:

real	0m6.599s
user	0m0.005s
sys	0m0.012s

If not on ProtonVPN, the connection remains open for much longer (I got impatient, so terminated with ^C):

H=kagi.com; time openssl s_client -4 -quiet -verify_return_error -servername $H $H:443
Connecting to 34.111.242.115
depth=2 C=US, O=Google Trust Services LLC, CN=GTS Root R1
verify return:1
depth=1 C=US, O=Google Trust Services, CN=WR3
verify return:1
depth=0 CN=kagi.com
verify return:1
^C
real	1m25.933s
user	0m0.016s
sys	0m0.007s

Nyaa

@yiwei-1 So can you please remove the Done tag, as this issue is still happening ?

yiwei-1

Nyaa

Are you still getting this error when you’re not using a VPN?

Nyaa

Not 100% sure, i almost always use a VPN (many domains are blocked on my work internet...)

But surely this should work over a VPN too? Proton is a popular VPN service, probably one of the top 5...
What would be needed to fix this?

mille

The problem seems to be as follows: every ProtonVPN server has a pool of addresses. Some or all of these addresses are marked by Google as high-risk, because a lot of simultaneous connections and/or attacks originate from them to Google properties. To prevent certain kind of attacks (Slowloris, resource depletion), Google tears down connections when the client is not sending data. Looking at a typical exchange with Wireshark, you will notice that if the model thinks more than 5 seconds, the connection will be torn down. If the model starts outputting its response before 5 seconds have elapsed, the connection will still be cut down exactly 5 seconds after the client sent its last TCP segment with a payload (i.e. the prompt). Although Wireshark shows my client sending many subsequent ACKs, none of these carry a payload, and they are thus not seen by Google as meaningful traffic:

{Google tearing down connection in the middle of model response even though it receives empty ACKs}

A workaround would be to try another egress IP address by switching to a different ProtonVPN server. Anecdotally, this seems to work, which suggests that Google has not put the entire ProtonVPN IP range on its naughty list.

Seeing that Kagi is using Google infrastructure, they might also be able to do something about this by increasing this 5s teardown rule for high-risk IP addresses in their Google console. I do not know whether that is (1) possible or (2) desirable, though, as it might open them up to potential DoS.

Nyaa

Fyi for anyone stumbling upon this issue, as it is still not fixed, a good workaround is to use for example the Claude 4.5 Sonnet Reasoning model;

Because it "streams" the "reasoning", the connection will stay alive, which prevents the issue from happening.