All About the Memory System

The Memory System automatically stores and retrieves important information from your adventure so the AI can stay consistent over long sessions.

It is built from two complementary features: Auto Summarization and the Memory Bank.

Why the Memory System Matters

Large language models have a limited context window. As your story grows, older text must be removed from the prompt, which can feel like the AI is forgetting important details.

The Memory System helps solve this by keeping both a high-level plot overview and relevant detailed memories, so your choices remain meaningful across long adventures.

How It Works: Core Overview

The system uses summarization models plus embedding-based retrieval. In simple terms, it continuously compresses story history and then fetches the most relevant memories when needed.

• Auto Summarization: Maintains a running plot-level summary.
• Memory Bank: Stores many smaller memory items and injects relevant ones dynamically.

Human Brain Analogy

The design mirrors two human memory strategies: compression (remembering key ideas, not every word) and retrieval (recalling details only when context makes them relevant).

In Questsmith terms, compression is summarization and retrieval is memory lookup from your Memory Bank.

What Is a Memory?

A memory is an AI-generated summary of a small chunk of earlier story actions. It is denser than normal prose and focused on key events, entities, and changes.

A new memory is created from six previous actions (including AI responses). The system starts creating memories after enough turns have accumulated, then continues on a repeating cycle.

Auto Summarization

Auto Summarization keeps a plot-level Story Summary current by periodically incorporating new events and compressing older details when needed.

This helps the AI preserve narrative direction while still leaving room for current scene context.

Story Summary Plot Component

Story Summary is a dedicated Plot Component. You can add and edit it manually, and if Auto Summarization is enabled, it updates automatically over time.

You can also monitor its token usage in the Context Viewer.

Editing and Existing Adventures

You can edit Story Summary manually. Your edits are included as input when future automatic summaries are generated.

If you make major changes far back in history, you may need to manually adjust the summary. Full historical re-summarization for every edit is constrained by model context and cost.

How the Memory Bank Retrieves Relevance

Each memory is converted into an embedding vector. The system compares vectors for your current context against stored memory vectors and ranks them by semantic similarity.

Top-ranked memories are inserted into the AI context as Used Memories. This lets the model recall specific details when they matter.

Memory Lifecycle

• Memory Bank starts empty in a new adventure.
• Memories accumulate as actions are summarized.
• When full, less-used memories are removed as Forgotten Memories.
• Frequently relevant memories can persist much longer.

How to Enable the Memory System

In the game settings sidebar, go to Gameplay -> AI Models -> Memory System, then enable Memory Bank and Auto Summarization.

Once enabled for one adventure, these settings remain active for future adventures until changed.

Membership Tier and Memory Bank Size

Memory System vs Plot Components

Plot Components are still important. AI Instructions, Story Summary, Plot Essentials, and Author's Note are always in context, while Story Cards are conditionally injected by keywords.

The Memory System reduces manual burden, but well-structured Plot Components still improve consistency and quality.

Best Practice Checklist

• Enable Auto Summarization early for best long-term quality.
• Use Story Summary edits to correct critical inaccuracies.
• Keep Plot Essentials concise and stable.
• Use Story Cards for broad world knowledge and keyword triggers.
• Review Context Viewer periodically to understand token allocation.