Skip to content

concepts.mdx

Latest commit

 

History

History
94 lines (68 loc) · 5.75 KB

concepts.mdx

File metadata and controls

94 lines (68 loc) · 5.75 KB
Raw
title Core Concepts
description The vocabulary of MemoryLake — and how individual and organizational memory fit together

MemoryLake models memory the way organizations actually work: knowledge has an owner, a scope, and a source. This page is the short orientation; each concept links to its deep dive.

You don't need to set any of this up by hand. When you sign up, MemoryLake automatically provisions a workspace and assistant for you — these concepts become useful as your usage grows from "my personal memory" to "my team's memory" to "memory inside my product."

The Hierarchy

Workspace                    ← top-level container and isolation boundary
├── Project                  ← a focused body of knowledge
│   ├── Documents            ← files whose content is parsed and indexed
│   ├── Conversations        ← message exchanges memory is extracted from
│   ├── Project Facts        ← knowledge scoped to this project
│   └── Open Data            ← subscribed external datasets
├── Actors                   ← who memories belong to (people and assistants)
│   └── Actor Facts          ← per-person memory, follows them across projects
└── Agents                   ← AI assistants operating in this workspace
Isolation boundaries, knowledge containers, and multi-tenant patterns. Identities, project facts vs. actor facts, and per-user memory. How documents and conversations become searchable memory.

Individual Memory and Organizational Memory

The same machinery serves two different kinds of knowledge, and knowing which you're building matters more than any API detail.

Individual memory Organizational memory
Belongs to A person, across every tool they use A team, in a shared workspace
Examples "Prefers TypeScript", "allergic to shellfish", "working on the marathon plan" "Refund window is 30 days", "Customer X renews in March", "we chose Postgres because…"
Lives as Actor facts (follow the person across projects) Project facts (scoped to the shared body of knowledge)
Governed by The person's own workspace — private Roles, fine-grained permissions, and per-resource scopes
The worry it solves "Will my tools remember me?" "Will knowledge survive people leaving?"

Both flow through the same pipeline — extraction, source traces, conflict checking, retrieval. The difference is scope and governance. See the Personal AI Memory and Team Knowledge Base scenarios.

Memory Hygiene: Traces and Conflicts

Two properties make memory trustworthy enough to build on:

  • Source traces — every memory links back to the conversation or document it came from. When an assistant asserts something, you can always answer "how do you know that?" Available in the console and via the trace API.
  • Conflict detection — when a new memory contradicts an existing one (a moved deadline, a changed policy), MemoryLake flags a conflict for review instead of silently keeping both. In teams, whoever holds the resolve permission decides what's true.

See Memories for how both work in practice.

Open Data

Open Data is curated external knowledge you subscribe a project to — research and academic sources, clinical trials, drug databases, financial markets, company fundamentals, economic indicators, patents, and more. Subscribed categories become searchable by agents alongside your own content, so a biotech project can cite PubMed while a fintech project cites SEC filings.

Boundary (Memory Router & MCP)

A boundary is the memory scope used by API traffic and MCP clients. It binds a workspace, project(s), and actors into a single reference:

  • With Memory Router, you pass it as ?boundary_id= on LLM calls — commonly one boundary per end user, so each user's conversations read and write only their own memory.
  • With MCP, you select the workspace and boundary on the OAuth consent screen when a client connects.

API Key

An API key (sk-…) is your credential for everything programmatic: the REST API, Model Router, and agent plugins. Create keys in the console under API Keys — each has a name and an expiry, and its permissions mirror those of the member who created it. Rotate or revoke anytime. See Authentication.

How the Concepts Map to You

You are… Workspace Projects Memory flows
An individual Your default personal workspace One per life/work context Your chats & files → your memories → every tool you connect
A team Shared team workspaces One per initiative or client Everyone's contributions → governed shared memory → every member's assistant
A product builder One workspace, or one per tenant Per product domain or per end user Your users' sessions → per-user memory via boundaries → your app's LLM calls
Experience memory in the console in five minutes. Deep-dive walkthroughs for personal, team, agent, and product use cases.