Delete Memories

Deletes multiple memories based on the provided parameters. You can delete memories by specifying an identifier (user_id or agent_id) and optionally memory_ids.

• If no memory_ids are provided, all memories associated with the given user_id or agent_id will be deleted.

• If scopes or metadata are included, the deletion will respect these filters — only memories matching the specified scopes and metadata will be removed.

• When memory_ids are provided, any IDs that do not belong to the specified user_id or agent_id will be ignored; only valid and related memories will be deleted.

Parameters

• memory_ids (list[str], optional) — List of memory ID.

• user_id (str, optional) — User identifier for the memory.

• agent_id (str, optional) — Agent identifier for the memory.

• scopes (set[MemoryScope], optional) — Specifies the target for memory management, either for a specific participant or for combined access (USER or ASSISTANT). Default is [USER, ASSISTANT].

  • USER: Writes and retrieves memories targeting messages from user.

  • ASSISTANT: Writes and retrieves memories targeting messages from assistant.

  • USER, ASSISTANT: Writes and retrieves memories targeting messages from both the user and assistant.

• metadata (dict[str, str], optional) — Extra metadata linked to the memory, which can be used to store any supplementary details or context related to it. Default is None.

Note: At least one of user_id or agent_id must be provided.

Returns

list[Chunk] — A list of deleted memory chunks. Each Chunk follows the GL SDK format and includes an id, content, metadata, and score.

Example

import asyncio
from gllm_memory import MemoryManager

# Initialize
memory_manager = MemoryManager()

memory_ids = [
    "e66f0832-2447-4c5b-ba5f-8fa85957f436",
]
result = asyncio.run(memory_manager.delete(
    memory_ids=memory_ids,
))

print(result)

or

import asyncio
from gllm_memory import MemoryManager
from gllm_memory.enums import MemoryScope

# Initialize
memory_manager = MemoryManager()

result = asyncio.run(memory_manager.delete(
    user_id="user_123",
    scopes=[MemoryScope.USER],
))

print(result)

Expected Output

[
    Chunk(
        id=e66f0832-2447-4c5b-ba5f-8fa85957f436,
        content=User's favorite programming language is Python,
        metadata={
            'message': 'Memory deleted successfully',
            'event': 'DELETE'
        },
        score=None
    )
]