#tried to DM you. Dont want to spam
1 messages · Page 1 of 1 (latest)
Grok found this for me - one sec
ok
@fleet edge is the hindsight dev thats here
Most Common Cause (and Fix)
This is a known bug reported in April 2026. The embedded daemon (launched automatically via uv cache) is flaky/buggy, especially with retain operations.
Best and easiest fix (what solved it for most people):
Create the config file to bypass the embedded daemon and force it to use the standalone Hermes service instead (which runs on port 8888):
mkdir -p ~/.hindsight
cat > ~/.hindsight/hermes.json << EOF
{
"hindsightApiUrl": "http://localhost:8888",
"bankId": "hermes",
"recallBudget": "high"
}
EOF
Then restarte gateway
So just changing the port in the config file ig? It's essentially the same as mine already but without the model I selected
Wait, I see another log file I did not know about in that directory. Will take a look
Looks like it hit a timeout and just, shutdown?
2026-04-21 22:03:23,907 - INFO - hindsight_api.daemon - Idle timeout reached (300s), shutting down daemon
INFO: Shutting down
INFO: Waiting for application shutdown.
Gonna see what hermes thinks
sometimes hermes dies and you gotta reach out and call a friend (grok 😉 )
heading to bed, gl with this - if you get it fixed let me know!
Haha yeah, thanks for the help!
Hey the daemon has a shutdown timeout if you don't use it but on the next usage it should restart - anyway today we are releasing 0.5.4 and we changed the default value to no timeout so I'd suggest to upgrade once it's out, I'll ping you here
Ok, yeah thats what I had come up with. Was it an issue with it just not auto starting on the next tool call like it should have?
@wicked stump can you upgrade to 0.5.4 ? there are 2 fixes related to this specific issue
Gonna try and update in a bit here after work 
Ok, bumped to the most recent version and have it running. Will wait a few hours and hopefully it doesn't shut down again, thank you!!
Awesome to hear you got it workign!
Yeah, up and running right now!