How memory works

How Flotira remembers you

Flotira keeps a long-term, private memory of your work — every conversation, every document, every fact you state — and pulls the right context into every reply. Here is exactly how that works, in plain English.

The problem with most AI

Most AI assistants forget you the moment the tab closes. You re-explain your business every session. You re-upload the same files. You repeat the same preferences. It’s exhausting — and it stops the AI from ever feeling like it actually knows you.

Flotira is built differently. Memory is not an add-on. It is the foundation.

What Flotira actually remembers

Memory isn’t a single bucket. Flotira keeps four kinds of memory, each one good at a different job. Together they make up your knowledge graph.

Second Brain

Every conversation + every document

A semantic archive of everything you’ve ever said to Flotira and every file you’ve shared. Searchable by meaning — ask for "that idea about pricing" and it finds the turn even if you never used those exact words.

Structured Facts

Specific things you’ve stated

When you say "my brand color is #ff6644" or "I ship on Thursdays," Flotira extracts that as a clean fact. No re-scanning old conversations — the answer is always one lookup away.

Wiki Pages

Synthesized project knowledge

As facts accumulate inside a project, Flotira quietly rolls related ones up into a wiki page — a living summary of everything the AI knows about that project. You can read it, edit it, or let it keep growing on its own.

Graph Links

Connections between memories

Every memory is linked to related ones — "this fact came from that conversation," "this wiki page absorbed those facts." When Flotira answers, it follows these links to pull in context you wouldn’t find with plain search.

Your knowledge graph

Every memory is linked to related ones. When you ask Flotira a question, it doesn’t just match keywords — it walks the graph and pulls in everything connected.

Memory System / Live

How Flotira remembers — every fact, traced.

How they connect

How recall works, step by step

Every time you send a message, Flotira runs this sequence before it even starts writing a reply.

  1. 1
    You ask a question
    Your message lands in the current project (or General Chat if there isn’t one).
  2. 2
    Flotira searches by meaning
    Your question is turned into a vector — a mathematical fingerprint of its meaning — and matched against your Second Brain. The top few most relevant conversations come back.
  3. 3
    It follows the graph
    From each matched conversation, Flotira walks the links one hop out — pulling in structured facts and wiki pages that are connected, even if they don’t mention your exact words.
  4. 4
    It trims to a budget
    Context is capped at roughly 2,500 tokens. Less important items get trimmed first so the most relevant memory always makes it into the reply.
  5. 5
    You get a grounded answer
    Flotira writes its reply with all of that context already in view — so it already knows who you are, what you’re working on, and what you’ve said before.

What happens when you change your mind

You tell Flotira your brand color is red. Three months later, you rebrand to blue. Here’s what happens:

  • Flotira notices the new value contradicts the old one
  • The old fact is marked as superseded — it stops influencing replies
  • The new fact takes over immediately
  • If the change is significant, it’s flagged in your Memory dashboard so you can confirm

The old version is kept in history, never deleted. If you ever want to undo, it’s still there.

In the Memory dashboard
brand_primary_color
Previous
#ff4444 (red)
Current
#3b82f6 (blue)
Mark Reviewed

Your memory belongs to you

Every memory is tied to your tenant and enforced at the database level with row-level security. Other Flotira users cannot see your memory. Support staff cannot browse your memory. Nothing you store is used to train anyone else’s model.

Project-scoped by default
Knowledge about a specific project stays inside that project. It doesn’t leak into unrelated chats.
You control what persists
Visit the Memory dashboard to review every fact, every conversation, and every wiki page — search, edit, or delete anything.
Exportable and portable
Your memory is your data. You can generate specs, wiki exports, or asset bundles at any time.

Memory vs. a context window — what’s the difference?

A context window is an AI’s short-term working memory — the text it can see while writing a single reply. It resets when the conversation ends. Typical windows today hold between 8,000 and 1,000,000 tokens. They don’t carry forward on their own.

Flotira’s memory is long-term storage — conversations, facts, documents, and summaries that live on after the tab closes. When you send a new message, Flotira searches this long-term memory and copies the relevant slice into the context window automatically.

Put simply: the context window is what the AI can see right now. Memory is what it can remember forever. Flotira connects the two so you never have to.

Frequently asked questions

The quick answers to the questions people ask most about Flotira’s memory.

Yes. Every chat, every document you share, and every fact you state is saved to your private long-term memory. Close the tab, come back in a month, and Flotira will still have the context.

An AI that actually knows you.

Stop re-explaining your business. Start working with an assistant that remembers.