mem_session_summary

Signature

mem.mem_session_summary(user_id: str, summary: str) -> SaveResult

Creates a type="summary" observation linked to the active session and persists it. Does NOT close the session — the session keeps running and last_activity_at is touched.

This is the right call when you want to checkpoint the conversation context (e.g., before a likely compaction) but the user is not done.

note

type="summary" is reserved — it cannot be passed to mem_save. This method is the only producer of summary observations.

Parameters

Parámetro	Tipo	Requerido	Descripción
user_id	str	sí	Owner of the active session.
summary	str	sí	Summary text persisted as a type="summary" observation. Must be non-empty. The observation title is auto-derived: summary[:80] if short enough, else summary[:77] + "...".

Returns

SaveResult (frozen Pydantic model):

• id (int) — observation id.
  • outcome (Literal["created", "updated", "deduped"]).
  • session_id (str) — the active session this summary belongs to.
  • topic_key (str | None) — always None for summaries.

Raises

• ValueError — summary is empty or whitespace-only.

Examples

Mid-session checkpoint

result = mem.mem_session_summary(
    user_id="alice",
    summary="Halfway through the auth refactor — login works, logout pending.",
)

print(result.id, result.outcome)   # e.g. 42, "created"
# session is still active — keep working

Structured checkpoint before compaction

checkpoint = """
## Goal
Refactor auth flow.

## Accomplished
- Login endpoint migrated to async.

## Next Steps
- Migrate logout + add CSRF tests.

## Relevant Files
- src/auth/routes.py
""".strip()

mem.mem_session_summary(user_id="alice", summary=checkpoint)

See also

mem_session_end

Concepts: Sessions