Changelog

Oct 20, 2025

Memory v0.4 Release

Memory v0.4 introduces a fully documented API for creating, reading, updating, deleting, and analyzing persistent AI memory. Developers can now directly manage long-term memory objects for each assistant, with immediate database and vector-store synchronization.

Available Endpoints

Function

Method

Endpoint

Description

Get All Memories

GET

/assistants/{assistant_id}/memories

Retrieves all memories for a specific assistant.

Add Memory

POST

/assistants/{assistant_id}/memories

Adds a new memory to the assistant. Supports content and optional metadata fields.

Get Memory by ID

GET

/assistants/{assistant_id}/memories/{memory_id}

Retrieves a specific memory object.

Update Memory

PUT

/assistants/{assistant_id}/memories/{memory_id}

Updates the content or metadata of a specific memory. Updates propagate to both DB and vector store.

Delete Memory

DELETE

/assistants/{assistant_id}/memories/{memory_id}

Permanently deletes a memory from the database and vector store.

Get Memory Stats

GET

/assistants/{assistant_id}/memories/stats

Returns statistics and limits related to the assistant’s memory usage.

Get Memory Operation Status

GET

/assistants/memories/operations/{operation_id}

Tracks the status of memory operations (add, update, delete) asynchronously.

What’s New in v0.4

  • Unified vector-store and database updates for every memory operation.

  • Enhanced metadata schema for richer contextual associations.

  • Operation-level tracking (/operations/{operation_id}) for large memory updates.

  • Streamlined error handling and validation across all memory routes.

Next Steps

Developers can explore the updated API reference at app.backboard.io/api/docs and start experimenting with Memory v0.4 via their assistant IDs.

<New Feature: Web Search Mode

AI Memory Should Be Portable, Not Owned by a Vendor>

Changelog