question(flaky): `cortex -v` hangs when checking if there is a newer release?

[freelerobot]

Flakey issue. Happens 1/3 times.

Steps:

1. cortex -v or any command really

❯ cortex-nightly -v
v0.5.0-70

2. Doesn't return, just hangs
3. After 10 seconds, we get "A new release is available"

What causes this?
Where's our endpoint for checking if there's a more recent version, and is this endpoint prone to lagging?

[vansangpfiev]

We check the new nightly version with https://delta.jan.ai endpoint. We are using the default timeout (10 seconds).
I think can set the timeout to 1 second, if connection timeouts, just not print any message on version update.

[freelerobot]

We check the new nightly version with https://delta.jan.ai endpoint. We are using the default timeout (10 seconds). I think can set the timeout to 1 second, if connection timeouts, just not print any message on version update.

Timeout to 1 s is a good workaround!

Q: I'm just wondering whether we need to check versions at the end of every command?

Rationale

1. Since we're hosting the endpoint ourselves, let's just assume it will be ddos'ed. Then will every cortex command, by default, takes 1 second, returning nothing? How will we build this mini-feature if we were building it for scale & to be antifragile to external dependencies?
2. Users typically execute many CLI commands in succession, during a single session with Cortex. If they don't want to update initially, then reminding them 10-20 times thereafter, is annoying.
3. Recall, Cortex is an on-prem, local tool. Many will use it on an intranet. So many external calls out is not our style.

Dumb Ideas

1. Reserve the version checks for just a few key commands, that users will almost always see:

cortex
cortex -v
cortex start                 # Server start is an impt entrypoint to using cortex, and will call attention.
cortex run                   # Starting an interactive chat shell is another impt entrypoint to using cortex

#Optionally
COMMAND -h            # Not annoying to put it in all the help commands as well

I'm not sure what the best fix is, but I do think shortening timeout to 1sec is a (good) but temporary patch to this issue.

cc @dan-homebrew

[vansangpfiev]

I would like to propose 2 approaches that can help to handle update check better:

• Approach 1: create a file to save the current version, latest version, last time check the update. This approach is simple, easy to implement, but we need to store our state which can be against our db-less/stateless design.
• Approach 2: server has a longer lifetime than the CLI, so we can let server checks the update instead of CLI. CLI only need to request local server for the new update information.

cc @dan-homebrew