docs: document message_store in /api/v1/health response [doc-updater]

[james-in-a-box]

Update docs/guides/deployment.md to reflect the components.message_store
field added to the /api/v1/health response in #3077 slice-6.

• Adds "message_store": {"status": "ok"} to the normal-response JSON example
• Documents when it reads "degraded" with reason MESSAGE_STORE_AUTO_FALLBACK_TO_MEMORY (unintentional auto→memory fallback at startup)
• Explains the distinction from explicit EGG_MESSAGE_STORE_BACKEND=memory (which is not flagged as degraded)
• Notes the mid-phase-restart message-loss risk that the flag surfaces (#3076)

Triggered by: #3144 (slice-6 of #3077)

Issue: none

Test Plan

• Automated: doc-only change, no code tests required
• Manual: verify the JSON example matches the actual /api/v1/health response shape in orchestrator/routes/health.py

Authored-by: egg

[egg-reviewer]

Approve. Doc-only change accurately documenting the components.message_store field added in #3077 slice-6. I verified every claim against the implementation rather than the PR description alone.

Verification against code

• JSON example shape and field order (docs/guides/deployment.md:390) — matches orchestrator/routes/health.py:159-168 exactly: state_store → state_store_summary → message_store → docker. The added "message_store": {"status": "ok"} line is placed in the correct position.
• {"status": "ok"} common case — matches health.py:144.
• Degraded shape {"status": "degraded", "reason": "MESSAGE_STORE_AUTO_FALLBACK_TO_MEMORY"} — matches health.py:138-142 and the marker constant MEMORY_FALLBACK_MARKER at message_store.py:620.
• Top-level status becomes "degraded" on auto-fallback — matches health.py:148 (healthy = state_store_healthy and not message_store_degraded) and :160.
• Auto-fallback trigger (EGG_MESSAGE_STORE_BACKEND unset/"auto" + Redis unreachable at startup) — matches message_store.py:696,711,721-730, which flips _memory_fallback_degraded = True only on the auto exception path (the redis-explicit path re-raises instead).
• Explicit EGG_MESSAGE_STORE_BACKEND=memory does NOT set degraded — matches message_store.py:698-709: that branch returns MessageStore() with a warning-level log and never touches the degraded flag. This is the most important distinction in the doc and it is correct.
• #3076 mid-phase-restart message-loss risk — consistent with the in-line code commentary at message_store.py:724-736.

The prose paragraph at :400 is placed directly after the state_store_summary description, mirroring the JSON field ordering, which keeps the section readable.

Non-blocking note

The "Normal response" JSON example omits process_start_time, last_unhealthy_at, and recent_transitions, which the real response includes (health.py:169-172). This is a pre-existing simplification — the example already trimmed those fields before this PR — so it is out of scope here, but worth tightening in a future pass if the example is meant to be exhaustive.

No security, correctness, or robustness concerns: this changes only Markdown, and the documented behavior faithfully reflects the shipped code.

— Authored by egg

[james-in-a-box]

egg review completed. View run logs

1 previous review(s) hidden.