Feedback API

Rate a memory's usefulness after you retrieve and use it. Positive feedback increases a memory's importance score over time; negative feedback decreases it. This signal feeds into the scheduled importance recalculation job. Importance affects which memories the session-context endpoint returns first, min_importance filtering, and pruning candidates; it does not change search result order.

See Session Lifecycle for how feedback integrates with the broader memory health cycle.

Base URL and Authentication

https://api.neuroloom.dev

All requests require an API key:

Authorization: Token $MEMORIES_API_TOKEN

Rate a Memory

POST /api/v1/memories/{id}/feedback

Record whether a memory was useful after you retrieved and acted on it. Call this after using (or deciding not to use) a retrieved memory to provide an explicit quality signal.

memory_rate(
  id="a1b2c3d4-e5f6-7890-abcd-ef1234567890",
  useful=true,
  context="This pattern resolved the N+1 query issue in the sessions router. The selectinload hint was exactly what was needed."
)

import httpx, os

response = httpx.post(
    "https://api.neuroloom.dev/api/v1/memories/a1b2c3d4-e5f6-7890-abcd-ef1234567890/feedback",
    headers={"Authorization": f"Token {os.environ['MEMORIES_API_TOKEN']}"},
    json={
        "useful": True,
        "context": "This pattern resolved the N+1 query issue in the sessions router.",
        "session_id": "sess-abc123",
    },
)
assert response.status_code == 201

curl -X POST "https://api.neuroloom.dev/api/v1/memories/a1b2c3d4-e5f6-7890-abcd-ef1234567890/feedback" \
  -H "Authorization: Token $MEMORIES_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "useful": true,
    "context": "This pattern resolved the N+1 query issue in the sessions router.",
    "session_id": "sess-abc123"
  }'

Path parameters

Request body

Response — 201 Created, the memory object the feedback was applied to.

{
  "id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
  "title": "Use selectinload for SQLAlchemy relationship loading",
  "importance_score": 0.9,
  "content_hash": null,
  "feedback_positive_count": 6,
  "feedback_negative_count": 1,
  ...
}

feedback_positive_count and feedback_negative_count reflect the updated totals after this submission.

How Feedback Affects Importance

Feedback is not applied to importance_score immediately. It is incorporated during the scheduled importance recalculation job that runs periodically across the workspace. After the job runs, memories with net positive feedback will have a higher importance_score. That raises their standing for session context, min_importance filters, and pruning, but does not move them up in search results, which are ordered by relevance to the query.

The recalculation formula weights:

• feedback_positive_count — increases importance
• feedback_negative_count — decreases importance
• access_count and retrieval_count — recency and frequency of use
• Original confidence_score — baseline trust in the memory

When to Submit Feedback

Submit positive feedback when a retrieved memory directly helped you complete a task or make a decision. Submit negative feedback when a retrieved memory was outdated, incorrect, or irrelevant to the context it appeared in.

The MCP memory_rate tool is designed for use at the end of a task, after you have evaluated whether the retrieved context was genuinely useful. If you used memory_search or memory_get_detail to retrieve a memory, call memory_rate once you know whether it helped.

Error Responses

{ "detail": "Memory '00000000-0000-0000-0000-000000000000' not found." }

Related

• Memory Lifecycle — how feedback signals feed into importance score recalculation
• What is a Memory — importance score fields and their meaning