fix: gracefully handle missing systemctl in container environments

[dashed]

Summary

hermes status and hermes gateway status crash with FileNotFoundError when running inside Docker/Kubernetes containers where systemd is not available.

This PR adds container/K8s environment detection and shows proper gateway status using PID-based health checks instead of systemctl.

Before (container)

$ hermes status
...
Traceback (most recent call last):
  File ".../status.py", line 293, in show_status
    result = subprocess.run(["systemctl", "--user", "is-active", ...])
FileNotFoundError: [Errno 2] No such file or directory: 'systemctl'

After (container)

◆ Gateway Service
  Status:       ✓ running
  Manager:      container/K8s

The output matches the systemd path's format exactly (✓ running / ✗ stopped) — the only difference is the manager line showing container/K8s instead of systemd (user).

Problem

On Linux, show_status() and systemd_status() unconditionally run systemctl --user is-active hermes-gateway. The except clause only catches TimeoutExpired, not FileNotFoundError — so containers without systemd get an uncaught exception traceback.

This affects any containerized deployment: Kubernetes pods, Docker containers, LXC, or any environment without systemd as the init system.

Fix

Container detection

New _is_container_env() helper in gateway.py checks:

• KUBERNETES_SERVICE_HOST env var (injected by K8s into every pod)
• shutil.which("systemctl") as fallback (catches Docker and other non-K8s containers)

Gateway status in containers

Instead of showing "N/A", the container path uses is_gateway_running() from gateway.status — a PID-based liveness check that works in any environment:

• Reads gateway.pid from HERMES_HOME
• Verifies the process is alive via os.kill(pid, 0)
• Returns True/False → displays ✓ running / ✗ stopped

For systemd_status() (the hermes gateway status command), the container path also shows runtime details from gateway_state.json when available (platform statuses, gateway state).

_is_service_running()

Returns False in container environments — correct because there is no systemd service (the gateway process IS running, but not as a systemd unit).

Changes

File	Change
hermes_cli/gateway.py	_is_container_env() helper; container-aware systemd_status() with runtime details; _is_service_running() early return
hermes_cli/status.py	Container detection + PID-based health check via is_gateway_running()
tests/hermes_cli/test_container_detection.py	10 tests (unit + e2e)

Behavior matrix

Environment	Before	After
Linux + systemd	✓ running / ✗ stopped	No change
macOS + launchd	Works normally	No change
K8s container	Crashes with FileNotFoundError	✓ running / ✗ stopped, Manager: container/K8s
Docker container	Crashes with FileNotFoundError	✓ running / ✗ stopped, Manager: container/K8s

Test coverage

10 tests in tests/hermes_cli/test_container_detection.py:

Class	Tests	What's tested
TestIsContainerEnv	4	K8s env var, missing systemctl, normal Linux, both indicators
TestSystemdStatusContainer	2	Running status in containers, no subprocess calls
TestIsServiceRunningContainer	2	Returns False in containers, calls systemctl otherwise
TestStatusOutputContainer	2	E2E: full show_status() output — container vs normal Linux

uv run pytest tests/hermes_cli/test_container_detection.py -v

Test plan

• All 10 tests pass
• hermes status shows ✓ running + container/K8s in K8s pod (verified on live deployment)
• hermes gateway status shows running state with runtime details in containers
• Non-container behavior completely unchanged
• Existing test suite unaffected (no regressions)

[teknium1]

Closing as duplicate — this issue is fixed in #8614 (salvaged from @alt-glitch's comprehensive #8573). Thanks for the contribution!