> For the complete documentation index, see [llms.txt](https://oodlz-ai-studio.gitbook.io/docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://oodlz-ai-studio.gitbook.io/docs/inbound-agent/the-brain/conversation-memory.md).

# Conversation Memory

10-message rolling memory per user, 24-hour TTL, answered without tool calls — the agent recalls previous turns instantly.

> **Memory questions answered from history, not from another tool call.**

Stateless chatbots forget everything the moment a conversation ends. Ask one "what did I just ask about" and it has no idea. The Oodlz inbound agent maintains a 10-message rolling context window per user in Redis with a 24-hour TTL — and answers memory questions directly from that history without firing a tool call.

## How It Works

Each turn — message in, reply out — appends to the user's session memory. The memory window holds the last 10 messages and rolls forward as new turns arrive. When the user references something from a previous turn ("the second one", "go back to that brand", "what did I just ask about"), the agent reads from session history and replies — no vector search, no FAQ lookup, no latency from a tool round trip.

The memory window expires 24 hours after the last turn. Within that window, the conversation is continuous. After it, the next message starts a fresh session.

The same memory store is keyed by the `{platform}_{senderId}` scheme used by every channel, so memory continuity holds whether the user stays on one channel or switches.

## What You Get

* Previous-turn references answered directly — no tool call, no latency, no cost on memory questions.
* 10-message rolling window — recent enough to stay relevant, deep enough to span a typical interaction.
* 24-hour TTL — generous for active sessions, automatic cleanup for abandoned ones.
* Same memory across channels — the rolling window doesn't reset when the user switches surface.