Electron app should be able to run when sidecar is already running

[BYK]

If there's already a sidecar instance running on the default port, Electron app should just connect to it and spawn its own server in the case the existing one dies for some reason.

This is in line with how we handle proxying with the --stdio-mcp option.

[MathurAditya724]

Are you suggesting we create a failover system for the sidecar? That is -

1. We check server health endpoint every hour
2. if fail, we have a retry system which tries to restart the server 3 times, increasing the interval of retry by a (interval * x)
3. If that too fail, we can show a dialog that we are unable to connect to the sidecar and some possible solutions

Is this summaries the task, or did I missed something?

[BYK]

This is the other way around:

If there's a sidecar instance already running on the default port, the electron app launch would probably fail as the sidecar won't come up, complaining about another server already running.

Instead of that, we should just skip running the server and let the UI connect to the existing server BUT, if the connection to that server errors we should try to (re)start our server after trying to connect a few times, ideally with exponential back off with no more than a total of 5 second wait (so can be 3 tries starting with 0.5 seconds for instance)

This should also cover the scenario you mentioned above.

[BYK]

Pseudo code

if not server_running:
  Start_server

if server_disconnected:
  Retry connection
  if waited more than 5 seconds:
    Try to start our server
      On error: crash and inform the user