feat: include operational health in /health endpoint

[chriscoey]

Problem

/health currently returns:

{"status":"healthy","database":"connected"}

This reports "healthy" even when a bank has failed operations, stuck consolidation, or other operational issues. I had 24 failed consolidation operations and 37 pending items — /health gave no indication. The implementation (memory_engine.py:health_check()) only runs SELECT 1 against the database.

Proposal

Include operational health signals in the response:

• consolidation_status: ok / degraded / stuck
• failed_operations_count: total across all banks
• pending_consolidation_count: total across all banks
• last_consolidation_at: timestamp of most recent successful consolidation

Example:

{
  "status": "healthy",
  "database": "connected",
  "consolidation_status": "degraded",
  "failed_operations": 24,
  "pending_consolidation": 37,
  "last_consolidation_at": "2026-03-29T01:57:06Z"
}

If adding fields to /health is undesirable (some monitoring systems expect a simple 200/503), a separate /health/detailed endpoint would also work.

Use case

Automated monitoring scripts and audit workflows that need to detect operational degradation without querying each bank's /stats individually.

[nicoloboschi]

Hi, this is a good idea. but you're looking for Bank Health.
You could have different configuration per-bank and some of them could be ok while others failing for some reason.
So maybe we could have /banks/{bank_id}/health endpoint.

What do you think?

[chriscoey]

Thanks for the quick turnaround on #745, #749, and #751 -- really appreciate the effort.

Per-bank health makes a lot of sense -- you're right that bank-level config differences mean global health isn't granular enough.

The fields that would help me most:

• failed_operations count
• pending_consolidation count
• last_consolidated_at timestamp
• Ideally some signal for whether consolidation is progressing vs stuck (pending count decreasing over time)

Right now /health returns "healthy" even when a bank has accumulated failed operations, so I end up piecing together bank stats separately to get the real picture. A /banks/{bank_id}/health endpoint that surfaced this would be much cleaner.