Advanced / Created or skipped

Write Outcomes

A successful write can create a memory or report that nothing was stored. Both are normal responses, and your code has to branch on them explicitly.

Two valid results

Compare the two write outcomes

{
  "id": 4821,
  "text": "The user prefers concise answers.",
  "tags": ["preference"],
  "source": "chat",
  "importance": 0.8,
  "is_summary": false,
  "created_at": "2026-08-01T12:04:11Z"
}

A record exists. Keep the id if you may need to update, summarize, or forget it later.

Client code

Branch on the outcome

from memorysync import AddSkippedResponse

result = ms.add("The user prefers concise answers.", tags=["preference"], source="chat")

if isinstance(result, AddSkippedResponse):
    log.info("not stored: %s", result.reason)   # expected, not an error
else:
    save_reference(result.id)                   # a record now exists

Treating a skipped write as a failure causes retry loops and misleading error rates. Count it as its own outcome instead.

Input quality

Give a write the best chance

Send one durable fact per call rather than a whole transcript.
Write statements that will still be useful next week, not conversational filler.
Attach source and tags so the record can be filtered and audited later.
Use bulk_add() for batches; each item reports its own created, skipped, or rejected status.

Copy and run

Copy a working example

from memorysync import AddSkippedResponse

result = ms.add("The user prefers concise answers.", tags=["preference"])

if isinstance(result, AddSkippedResponse):
    print("skipped:", result.reason)
else:
    print("created:", result.id)

Was this page helpful?

← Previous

Advanced Overview

Importance & Ranking