Keep the story. Lose the context bloat.
Memory Books is a SillyTavern extension for creating and managing long-term chat memory. It turns completed scenes or automatically selected portions of a chat into structured, editable lorebook entries that preserve important events, decisions, relationships, discoveries, promises, and unresolved threads.
As your chat grows, older messages can be hidden from the active context while their important information remains available through the Memory Book. You keep a durable record of what actually happened without carrying the entire raw conversation inside every new request.
Use Memory Books manually when you want precise control over scene boundaries, or automate the process and let it create memories as the chat continues.
Not sure how AI memory actually works? Read How Memory Works in SillyTavern for a plain-language explanation of chat history, summaries, lorebooks, vector retrieval, trackers, and where Memory Books fits.
STMB has a lot of features, and its documentation became too large to navigate comfortably as a traditional README. Choose whichever guide option works best for you.
Just want to get STMB up and running? Start here. (I highly recommend you come back for the interactive guide soon though!)
For users willing to sign in with a Google account:
Use the Interactive Memory Books Guide to ask questions about installation, configuration, features, workflows, and troubleshooting.
The guide has the Memory Books documentation available as reference and can walk you through the extension interactively.
For users who do not use Google—or who prefer another AI assistant:
Download the Memory Books AI Reference Manual, upload it to the AI assistant of your choice, and ask it questions about Memory Books.
The manual is a dense, consolidated source of truth designed specifically for AI reference. It covers installation, ordinary and advanced workflows, settings, terminology, troubleshooting, and current feature behavior.
SillyTavern Memory Books is Copyright © 2024–2026 Aiko Hanasaki.
The original code in this repository is licensed under the GNU Affero General Public License v3.0. Modified versions and forks must preserve applicable copyright and license notices, identify their modifications, and comply with the AGPL's source-availability requirements.
See LICENSE.