Documentation
Continuity Engine

Intelligent Context Preservation

Never lose context in long conversations. The Continuity Engine automatically extracts, summarizes, and preserves the most important information from your chats.

What is the Continuity Engine?

When conversations grow long, AI assistants typically forget earlier context due to token limits. The Continuity Engine solves this by intelligently extracting and preserving key information - goals, decisions, requirements, and important references - so the AI never loses track of what matters.

Think of it as giving the AI a "working memory" that persists across the entire conversation, no matter how long it gets.

Rolling Summaries

Automatically extracts goals, decisions, and requirements from your conversation.

Pinned Memory

Pin important messages to ensure they're never forgotten or compressed.

Auto-Compaction

Intelligently compresses older messages while preserving essential context.

User Memory

Store persistent facts, preferences, and constraints across all conversations.

Three-Tier Context System

The Continuity Engine uses a three-tier architecture to manage context efficiently:

Tier 1: Raw Messages

The most recent messages (typically last 20) are kept in full detail. These provide immediate context and are never compressed.

Tier 2: Rolling Summary

Older messages are distilled into a structured summary containing goals, decisions, requirements, open questions, and important references. This preserves meaning while dramatically reducing token count.

Tier 3: Pinned Memory

Critical information that you explicitly pin is preserved verbatim and never compressed. Perfect for key decisions, important code snippets, or crucial requirements.

Context Health Indicator

The circular indicator in your chat header shows the current context preservation ratio - a measure of how much original context has been retained:

RatioColorStatus
70-100%GreenHealthy - minimal compression
40-69%AmberModerate - some summarization applied
<40%RedHeavy compression - consider pinning key items

Pro Tip

Click the context indicator to open the Summary Drawer, where you can view extracted context, manage pinned items, and manually trigger compaction.

Key Features

Automatic Context Extraction

Goals, decisions, requirements, and references are automatically identified and added to the rolling summary.

Message Pinning

Pin any message to ensure its content is never compressed or forgotten. Perfect for critical decisions and important code.

Smart Compaction

Three levels of compaction (light, deep, aggressive) allow fine-tuned control over how much context is preserved vs. token savings.

User Memory

Store persistent preferences, facts, constraints, and definitions that apply across all your conversations.

Message Exclusion

Mark specific messages to be excluded from the rolling summary, useful for off-topic discussions or sensitive content.

Learn More