Update Memories

Modify existing memory content or metadata in Cognis

The update method modifies an existing memory record’s content or metadata. It returns the updated CognisMemoryRecord after the change is applied.

update() is a hosted-only method. The open-source lyzr-cognis package handles updates automatically — when you add() messages containing updated facts, the extraction pipeline deduplicates and updates existing memories based on similarity thresholds.

Basic Usage

from lyzr import Cognis

cog = Cognis(api_key="sk-your-api-key")

updated = cog.update(
    memory_id="mem_abc123",
    content="User's name is Alice Johnson (updated)",
)

print(updated.content)

Method Signature

cog.update(
    memory_id: str,
    content: str | None = None,
    metadata: Dict[str, Any] | None = None,
    owner_id: str | None = None,
) -> CognisMemoryRecord

Parameters

Parameter	Type	Required	Description
memory_id	str	Yes	The ID of the memory to update.
content	str	No	New content to replace the existing memory text.
metadata	Dict[str, Any]	No	New metadata to attach to the memory.
owner_id	str	No	Owner identifier for scoping.

Response

Returns the updated CognisMemoryRecord with all fields reflecting the changes:

{
  "id": "mem_abc123",
  "content": "User's name is Alice Johnson (updated)",
  "owner_id": "user_alice",
  "version": 2,
  "is_current": true,
  "updated_at": "2025-01-15T12:00:00Z"
}

Update Content

# Search for the memory first
results = cog.search(query="user's name", owner_id="user_alice")
memory_id = results[0].id

# Update the content
updated = cog.update(
    memory_id=memory_id,
    content="User's name is Alice Johnson, she recently got married.",
)

print(f"Updated: {updated.content}")
print(f"Version: {updated.version}")

Update Metadata

updated = cog.update(
    memory_id="mem_abc123",
    metadata={"category": "personal", "confidence": 0.95},
)

print(updated.metadata)
# {"category": "personal", "confidence": 0.95}

Update Both Content and Metadata

updated = cog.update(
    memory_id="mem_abc123",
    content="User prefers Python 3.12 for new projects.",
    metadata={"topic": "programming", "last_verified": "2025-01-15"},
)

Search, Update, and Verify Workflow

A common pattern is to search for a memory, update it, and verify the change:

# 1. Search for relevant memory
results = cog.search(query="programming language", owner_id="user_alice")

if results:
    memory_id = results[0].id
    print(f"Found: {results[0].content}")

    # 2. Update it
    updated = cog.update(
        memory_id=memory_id,
        content="User is proficient in Python, TypeScript, and Rust.",
    )

    # 3. Verify
    verified = cog.get_memory(memory_id=memory_id)
    print(f"Verified: {verified.content}")

Async Usage

import asyncio
from lyzr import Cognis

async def main():
    async with Cognis(api_key="sk-your-api-key") as cog:
        updated = await cog.aupdate(
            memory_id="mem_abc123",
            content="Updated memory content",
            metadata={"source": "manual_update"},
        )
        print(updated.content)

asyncio.run(main())

Use include_historical=True in cog.get() to retrieve previous versions of a memory after updating it. This lets you track how a memory has changed over time.