BerriAI/litellm · error · HTTPException

Memory with key '{key}' already exists.

Error message

Memory with key '{key}' already exists.

What it means

HTTPException raised on PUT when the initial create attempt lost a race — a concurrent request created the row with the same key after the existence check, and the retry/fallback also found a conflict. Signals the key now exists so the caller should retry as an update.

Source

Thrown at litellm/proxy/memory/memory_endpoints.py:537

                "team_id": team_id,
                "created_by": user_api_key_dict.user_id,
                "updated_by": user_api_key_dict.user_id,
            }
            if body.metadata is not None:
                create_data["metadata"] = _serialize_metadata_for_prisma(body.metadata)
            try:
                row = await _memory_table(prisma_client).create(data=create_data)
            except Exception as e:
                # Race: a concurrent PUT/POST created the row after our check.
                # Re-read and fall back to an update so the PUT stays idempotent
                # instead of surfacing a 500 on a unique-violation.
                if not _is_unique_violation(e):
                    raise
                existing_after_race: Final = await _find_existing()
                if existing_after_race is None:
                    # Row exists globally but isn't visible to this caller
                    # (owned by someone else). Treat as conflict.
                    raise HTTPException(
                        status_code=409,
                        detail=f"Memory with key '{key}' already exists.",
                    )
                # Same write-authorization check as the non-race path.
                await _assert_write_access(prisma_client, existing_after_race, user_api_key_dict)
                row = await _memory_table(prisma_client).update(
                    where={"memory_id": existing_after_race.memory_id},
                    data=data,
                )
    except HTTPException:
        raise
    except Exception as e:
        raise _internal_error("Error upserting memory: %s", e, "Internal error updating memory entry.")

    return _row_to_model(row)


@router.delete(

View on GitHub (pinned to 77b7c6c40c)

Solutions

1. Choose a different key, or update the existing memory entry instead of creating a new one.

When it happens

Trigger: Thrown at litellm/proxy/memory/memory_endpoints.py:537 when the library encounters an invalid state.

Common situations: See trigger scenarios.

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/d2be948b5d27203c. Report an issue: GitHub.