Hello,

We do not support HTTP/2. HTTP/2 is accepted, but it is demuxed down to HTTP/1.1 when our proxy connects to your application.

Thanks Brody, how do I enable persistent connection beyond 60s?

Sorry I am super new to deploying app so might be some dumb questions

I use Groq in production. I'm deploying the proxy on Railway and embedding my Groq LLM API key into the proxy so my client application can request through my proxy. I've noticed it's hard to keep the connection longer than 60 seconds. Sending "Keep Alive" doesn't keep it. Is there any way to work around this?

My TCP/TLS latency are quite high

(base) ZhaobangJetWu (main *) ~
$ curl -w @- -o /dev/null -s "https://clio-backend-production.up.railway.app/" <<<'timenamelookup: %{timenamelookup}\ntimeconnect: %{timeconnect}\ntimeappconnect: %{timeappconnect}\ntimetotal: %{timetotal}\n'
timenamelookup: 0.778054 timeconnect: 0.779268
timeappconnect: 1.415058 timetotal: 1.507795

(base) ZhaobangJetWu (main *) ~
$ curl -w @- -o /dev/null -s "https://clio-backend-production.up.railway.app/" <<<'timenamelookup: %{timenamelookup}\ntimeconnect: %{timeconnect}\ntimeappconnect: %{timeappconnect}\ntimetotal: %{timetotal}\n'
timenamelookup: 0.005307 timeconnect: 0.008736
timeappconnect: 0.630628 timetotal: 0.718160
(base) ZhaobangJetWu (main *) ~

$ curl -w @- -o /dev/null -s "https://clio-backend-production.up.railway.app/" <<<'timenamelookup: %{timenamelookup}\ntimeconnect: %{timeconnect}\ntimeappconnect: %{timeappconnect}\ntimetotal: %{timetotal}\n'
timenamelookup: 0.003124 timeconnect: 0.004879
timeappconnect: 0.621428 timetotal: 0.701561
(base) ZhaobangJetWu (main *) ~
$

On our side, we allow for persistent connections for up to a maximum of 15 minutes, so anything stopping your connections at a timeframe shorter than that would be something on your side.

I will go ahead and place a bounty on this thread in hopes that the community can help you debug where you have this 60-second timeout set.