Description

The /update command triggered from Telegram (or any non-TTY context) fails with exit code 1 when the update introduces new config options that trigger the migration prompt.

Root cause

In hermes_cli/main.py, the cmd_update() function calls input("Would you like to configure them now? [Y/n]: ") unconditionally (around line 2819). When /update is triggered from Telegram, the update process is spawned via nohup bash -c ... (or systemd-run --scope) with no TTY attached. input() receives EOF immediately, which causes the function to either crash or exit with code 1 before reaching the gateway restart logic.

Impact

• The git pull and dependency install succeed, so the code is updated on disk
• But the gateway never restarts, so the old code keeps running in memory
• The user gets a failure notification ("update failed") even though the code was partially updated
• The next /update from Telegram reports "already up to date" since the pull already landed

Reproduction

1. Set up Hermes with Telegram as the control channel and launchd as the service manager
2. Wait for a release that introduces new config options (triggering needs_migration)
3. Send /update via Telegram
4. Observe exit code 1 and gateway not restarted

Suggested fix

Guard the input() call with a TTY check:

if sys.stdin.isatty():
    response = input("Would you like to configure them now? [Y/n]: ").strip().lower()
else:
    print("Non-interactive session — skipping config migration.")
    print("Run 'hermes config migrate' later to configure.")
    response = "n"

This allows the update to complete and restart the gateway in non-interactive contexts, while preserving the interactive prompt for terminal users.

Environment

• Hermes v0.4.0 (v2026.3.23+)
• macOS (Darwin), Mac mini M1
• Gateway managed by launchd
• Telegram as primary control channel