CLI should detect and handle stale pre-rename gateway processes

[yourfriendfitz]

Problem

After the engie → familiar rename (commit 8a1e743), running familiar with no arguments fails with:

Failed to connect to gateway: Gateway auth failed: {"code":"INVALID_REQUEST","message":"invalid connect params: at /client/id: must be equal to constant; at /client/id: must match a schema in anyOf"}

This happens because the old gateway process (clawdbot-gateway) is still running from before the rename. It validates against a JSON Schema that expects the old client ID (engie-ui / clawdbot-ui), but the updated CLI now sends familiar-ui.

Expected Behavior

The CLI (or familiar doctor) should detect stale gateway processes running under old names and either:

1. Warn the user that an old gateway is running and needs to be restarted
2. Automatically kill the old process and start the new one

Suggested Implementation

• In familiar doctor and/or the chat connect flow, scan for processes matching old gateway names (clawdbot-gateway, engie-gateway, etc.)
• If found, surface a clear message: "Found a stale gateway process (clawdbot-gateway, PID 1553). Run familiar stop && familiar start or kill it manually."
• Consider adding a --force-restart flag to familiar start that kills any old-named gateway processes before starting
• familiar status could also flag when the running gateway PID doesn't match the expected process name

Context

Old process names to check for:

• clawdbot-gateway
• engie-gateway

Current expected process name should be derived from the updated codebase.

🤖