Memory

How Rian retains and uses context.

Session memory

Within a single conversation, Rian remembers everything said. Customers do not need to repeat themselves.

Long-term memory

When a contact is identified (via Beacon.identify() or a previous login), Rian can reference previous conversations and stored contact attributes such as plan, company, and external ID.

Memory tiers

Rian uses a four-tier memory architecture visible in the Sandbox Memory tab:

TierDescription
HotActive session context; lowest latency
WorkingShort-term working memory for the current flow
WarmRecently accessed knowledge; cached for speed
ColdFull knowledge base retrieval; highest latency

Knowledge base

Rian uses your Storage content to answer questions. Add content from Storage in the sidebar:

SourceHow to add
DocumentUpload PDF, DOCX, or plain text from the Upload Knowledge tab
WebsiteEnter a URL in the Index Website field; Renprofile crawls and indexes it
ArticleWrite directly in the article editor within a category

AI-generated conversation summary

After each conversation, Rian generates a summary stored in the Agent notes field of the Conversation Details panel. Agents can read and edit this summary.

On this page