optimize.health

Read-only health diagnostic for the workspace memory store. Returns a quality score on a 0–100 scale plus underlying signal counts.

GET /v1/optimize/health

Free diagnostic — does not count against your plan’s memory-write or memory-retrieve quotas. Suspended and limit-reached workspaces can still call this endpoint.

Request

curl https://api.mnemexa.com/v1/optimize/health \
  -H "Authorization: Bearer mnx_ws_YOUR_KEY"

No request body. Workspace is derived from the API key.

Response

{
  "workspace_id": 42,
  "quality_score": 87,
  "signals": {
    "total_memories": 1240,
    "stale_count": 42,
    "stale_rate": 0.0339,
    "never_retrieved_count": 18,
    "never_retrieved_rate": 0.0145,
    "duplicate_count": 0,
    "duplicate_rate": 0.0,
    "overlong_count": 3,
    "overlong_rate": 0.0024,
    "optimized_count": 91,
    "optimized_rate": 0.0734
  },
  "open_recommendations": 2,
  "last_run_at": "2026-05-15T06:00:00Z",
  "top_issue_types": [
    {"type": "stale", "count": 42},
    {"type": "overlong", "count": 3}
  ],
  "window_days": 30
}

Fields

Field	Type	Description
`workspace_id`	integer	The workspace being reported on.
`quality_score`	integer	Composite health score, 0–100. Higher is better.
`signals.total_memories`	integer	Total non-archived memories in the workspace.
`signals.stale_count`	integer	Memories not retrieved within `window_days` and considered candidates for decay.
`signals.stale_rate`	number	`stale_count / total_memories`.
`signals.never_retrieved_count`	integer	Memories that have never been retrieved since storage.
`signals.duplicate_count`	integer	Near-duplicate clusters detected (post-dedup-pipeline residuals).
`signals.overlong_count`	integer	Memories whose text exceeds the recommended length budget.
`signals.optimized_count`	integer	Memories that have been auto-improved (compressed, deduplicated, or merged).
`open_recommendations`	integer	Pending optimization recommendations (curated via the dashboard).
`last_run_at`	string \| null	ISO 8601 timestamp of the last optimization sweep.
`top_issue_types`	array	Up to three most-frequent signal types currently flagged.
`window_days`	integer	Time window used for “stale” / “recent” classification. Default 30.

See Self-Optimization for what each signal means and how the score is computed.

Errors

Status	Cause
401	Missing or invalid `Authorization` header.
403	Workspace deleted or client account suspended.

This endpoint never returns 429 for unit limits, and never 422 — there are no inputs to validate.

SDKs

import mnemexa

client = mnemexa.Client()
health = client.optimize.health()
print(f"Quality: {health.quality_score}/100")
print(f"Stale: {health.signals.stale_count} of {health.signals.total_memories}")

The MCP adapter exposes this as brain.health. See Python SDK and MCP Adapter.

← Previous memory.retrieve Next status →