ci: add runtime smoke test for live agent startup

[benvinegar]

Summary

Add a runtime smoke integration check that verifies baudbot can boot and serve session-control RPCs on real droplets.

What changed

• Added bin/ci/smoke-agent-runtime.sh

  • Starts baudbot
  • Waits for control-agent session-control alias/socket
  • Probes Unix-socket RPC (get_message) as baudbot_agent
  • Verifies bridge supervisor status artifact exists
  • Enforces short stabilization window and status health check
  • Stops baudbot cleanly (with trap-based cleanup)

• Updated droplet setup scripts to run runtime smoke:

  • bin/ci/setup-ubuntu.sh
  • bin/ci/setup-arch.sh

Validation

Local:

• bash -n bin/ci/smoke-agent-runtime.sh bin/ci/setup-ubuntu.sh bin/ci/setup-arch.sh
• npm run test:shell

Runtime:

• Manually executed full Ubuntu droplet flow with this branch; runtime smoke phase passed end-to-end.

Notes

• This is a liveness/runtime-health smoke test; it does not trigger an LLM turn.
• Follow-up for optional gated inference smoke tracked separately (TODO-583a50bc).

[greptile-apps]

Greptile Summary

Added runtime smoke test that verifies baudbot can boot, serve session-control RPCs, and remain stable on real droplets.

Key changes:

• New smoke-agent-runtime.sh script starts baudbot, waits for control socket, probes Unix-socket RPC, validates bridge supervisor artifact, enforces 20s stabilization window, and stops cleanly with trap-based cleanup
• Integrated runtime smoke test into both Ubuntu and Arch droplet setup scripts after CLI smoke tests
• Script uses proper bash error handling (set -Eeuo pipefail) and timeout mechanisms

Issues found:

• Python socket in RPC probe lacks explicit close() call (minor resource leak in short-lived smoke test)

Confidence Score: 4/5

• Safe to merge with one minor style improvement recommended
• Well-structured smoke test with proper error handling, timeout mechanisms, and cleanup logic. The only issue is a minor resource leak in the Python socket code that won't affect the short-lived test execution. Changes to setup scripts are minimal and safe.
• bin/ci/smoke-agent-runtime.sh has a minor socket cleanup issue but is otherwise well-implemented

Important Files Changed

Filename	Overview
bin/ci/smoke-agent-runtime.sh	New runtime smoke test script that starts baudbot, verifies socket creation, probes RPC, and checks stability. Has one potential issue with Python socket cleanup.
bin/ci/setup-ubuntu.sh	Added runtime smoke test call after CLI smoke tests - minimal change, straightforward integration.
bin/ci/setup-arch.sh	Added runtime smoke test call after CLI smoke tests - minimal change, straightforward integration.

_{Last reviewed commit: 903afd3}

[greptile-apps]

_{3 files reviewed, 1 comment}

_{Edit Code Review Agent Settings | Greptile}

[greptile-apps]

Socket not closed after use; could leak file descriptor

Suggested change

	client = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
	client.settimeout(4)
	client = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
	try:
	client.settimeout(4)

Prompt To Fix With AI

This is a comment left during a code review.
Path: bin/ci/smoke-agent-runtime.sh
Line: 79-80

Comment:
Socket not closed after use; could leak file descriptor

```suggestion
client = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
try:
    client.settimeout(4)
```

How can I resolve this? If you propose a fix, please make it concise.

[benvinegar]

Good catch — I wrapped the RPC socket usage in a try/finally and now always call client.close() to avoid leaking the descriptor.

_{Responded by dev-agent-modem-a8b7b331 using openai/gpt-5-mini.}