Skip to documentation
Let’s build
PUBLIC TECHNICAL PREVIEW Production access and integration details are agreed during onboarding.
Amora Docs/Persistent memory

DEVELOPER DOCUMENTATION

Persistent memory

Build continuity with profiles, summaries and retrieval.

Persistent memory is the application layer that gives a character continuity across sessions. A stored document collection alone does not provide a complete personal memory system.

Three kinds of context

ContextTypical purpose
Recent conversationThe immediate exchange and unresolved topics
Durable user profileExplicit preferences, names and approved personal context
Character knowledgeBackground information the character can retrieve

Keep each source identifiable so the application can update or remove the right information.

  1. Resolve the authenticated user, workspace and character.
  2. Load recent messages and permitted profile facts.
  3. Retrieve relevant character knowledge when needed.
  4. Generate the response with that bounded context.
  5. Summarize or analyze the completed exchange.
  6. Save approved updates through the application's memory policy.

Isolation and control

Scope memory by workspace, user and character. A user's private conversation must not become another user's context. Support correction, deletion and retention settings in your product design.

Avoid treating every generated assumption as a verified user fact. Preserve unknown values until the conversation supplies usable information.

Configuration service

We help define profile fields, summary behavior, retrieval rules, storage and lifecycle settings for your deployment. Persistence, access controls and retention are implemented and verified as part of that project scope.

Discuss a memory configuration.

Building something personal?

Talk to our team about your architecture, character and launch.

Discuss your integration ↗