Adaptive Retrieval-Augmented Generation Memory

Today just4o.chat uses one current memory system: Adaptive Retrieval-Augmented Generation Memory. Instead of pasting a full memory list every turn, the app retrieves the most relevant memory chunks for the active scope and injects that smaller memory context into the chat.

You still keep editable memories per scope, with entries up to 200 words and tier-based caps of 15 on Explorer, 30 on Builder, 75 on Studio, 100 on Aurora, and 200 on Pro.

Each scope has a main memory toggle. In a project, that project-memory toggle is the umbrella switch for whether the project’s memory participates at all.

Under that umbrella, Adaptive Memory: Read and Adaptive Memory: Write appear as separate foreground tool toggles. If the scope memory toggle is off, those foreground tool toggles are unavailable for that scope.

Adaptive Memory: Read lets the model search memories, files, and chats in the active scope when it needs more grounding. Adaptive Memory: Write lets it create, update, add notes to, or delete memories in that same scope while the conversation is happening.

This gives smarter models a much more agent-like way to work, similar to coding agents that iteratively inspect context before acting. The tradeoff is usage: more retrieval and extra tool rounds can make these chats cost more than a plain response that never uses those tools.

When a foreground write tool changes memory, the reply can surface badges like “Memory Created”, “Memory Updated”, or “Memory Updates” so you can review what changed.

The older Background Memory Agent path is kept for legacy context, but it is not the primary system to use when explaining today’s product behavior.

If you are documenting just4o.chat as it works now, lead with Adaptive Retrieval-Augmented Generation Memory and foreground model reads/writes. Treat the Background Memory Agent as the older background model, not the main runtime.