> ## Documentation Index
> Fetch the complete documentation index at: https://docs.autosage.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Long conversations

> How AutoSage keeps lengthy chats fast, coherent, and true to the details that matter.

AutoSage keeps long conversations running smoothly. As a chat grows large, it automatically summarizes the older parts of the conversation while keeping recent messages verbatim. This process — called **compaction** — lets the assistant stay consistent and responsive without re-reading the entire history on every message.

## What compaction preserves

Compaction is designed to hold onto the details that matter over a long conversation:

<CardGroup cols={2}>
  <Card title="Preferences" icon="user-gear">
    How you like things done, carried forward as the chat grows.
  </Card>

  <Card title="Decisions" icon="circle-check">
    Choices you've made earlier in the conversation.
  </Card>

  <Card title="Constraints" icon="ruler">
    Requirements and boundaries the assistant should keep respecting.
  </Card>

  <Card title="Open tasks" icon="list-check">
    Work still in progress that the summary identifies as relevant.
  </Card>
</CardGroup>

Recent messages stay exactly as written, while older context is distilled into a summary. This provides best-effort continuity, but a summary can omit details; restate critical constraints when they matter.

<Info>
  Compaction happens **automatically and transparently**. There's nothing to configure and nothing to turn on — long chats simply keep working.
</Info>

## Compaction vs. Memory

Compaction and [Memory](/memory/overview) both help the assistant remember, but they work at different scopes:

|         | Compaction                         | Memory                                   |
| ------- | ---------------------------------- | ---------------------------------------- |
| Scope   | A single long conversation         | Across separate conversations            |
| Purpose | Keep one chat coherent as it grows | Carry durable facts between chats        |
| Setup   | Automatic                          | Configured per tenant and knowledge base |

<Card title="Learn about Memory" icon="brain" href="/memory/overview">
  How durable facts carry across different conversations.
</Card>

## Getting the best results

<Tip>
  When you move on to a **genuinely new topic**, start a new chat rather than continuing a long, unrelated one. A fresh chat keeps answers focused on the task at hand — and costs less than carrying a long history that no longer applies.
</Tip>

## Next steps

<CardGroup cols={2}>
  <Card title="Chat overview" icon="comments" href="/chat/overview">
    How grounded, cited answers come together.
  </Card>

  <Card title="Memory" icon="brain" href="/memory/overview">
    Carry durable facts across separate conversations.
  </Card>

  <Card title="Quick vs Deep modes" icon="gauge-high" href="/chat/modes">
    Choose how deeply each message is answered.
  </Card>

  <Card title="Sharing a chat" icon="share-nodes" href="/chat/sharing">
    Share a read-only transcript of a conversation.
  </Card>
</CardGroup>