In short
One short definition per term, grouped by domain, each pointing at the page that develops it. Two vocabularies meet here: Linkr’s own — workspace, entity, cohort — and that of health data — OMOP, concept, long format.
The notions, not just the words
To understand OMOP, terminologies or the long format, the articles under Resources go further than definitions. This page is for settling a doubt while reading, not for learning a field.
Organising
Workspace — The container a team shares: its projects, its databases, its common resources. This is the level at which you are a member and hold a role. See Workspaces and projects.
Project — A study, a monitor, a question: its cohorts, analyses, dashboards and code. It belongs to a workspace and inherits its permissions.
Entity — Everything that exports, versions and publishes the same way: project, workspace, schema, database, ETL pipeline, concept mapping, script collection, quality rules, data catalog, plugin. Ten types, one shape. See Entities and sharing.
Lineage identity — The identifier that follows an entity from one installation to another. It is what allows a modified entity coming back to be recognised instead of duplicated.
Role — A set of rights, described by resource and action. Granted on a workspace, inherited by its projects, overridable project by project. See Authentication and permissions.
The data
Warehouse — The read-only side: the source databases, as they are. You query them, you do not write to them.
Lab — The side where you produce: datasets, code, dashboards. This is where results are made.
Database — A data source declared in a workspace: a file, a folder of files, or a connection to a server. See Databases.
Schema — A source’s description: its tables, its columns, and how Linkr should read them. A schema preset is a reusable schema, OMOP or MIMIC for instance. See Schemas.
Long format — One row per measurement: a patient, a variable, a value, a date. This is the shape of warehouses, made to hold any variable without changing structure.
Wide format — One row per unit of analysis — a patient, a stay, a day in hospital — and one column per variable. This is the shape an analysis expects.
Dataset — A table in wide format, produced by a script or imported. See Datasets.
ETL pipeline — The transformation that fills one database from another: this is how a home-grown model becomes OMOP. See ETL pipelines.
Concepts and terminologies
Concept — An entry in the database’s dictionary: a code and its label. Depending on the database, that dictionary is called concept, d_items, or carries a home-grown name. See Concepts.
Concept list — A selection you compose, belonging to a project, travelling with it.
Concept set — A dictionary imported into the workspace, read-only, that projects consult.
OMOP — A common data model for health research: the same tables and columns from one institution to the next, which makes an analysis replayable elsewhere.
Concept mapping — The work of matching your local codes to a standard vocabulary. See Concept mapping.
AI suggestion — A mapping proposal left by an agent, with its equivalence and a comment. It waits for your review and only becomes a mapping if you validate it. See Suggestions.
Selecting and analysing
Cohort — A population defined by inclusion and exclusion criteria, expressed without writing a query. See Cohorts.
Attrition — The count of patients removed at each criterion. It is what makes a selection legible and publishable.
Unit of analysis — What one row of a dataset represents: a patient, a stay, a day. The choice is made when building the table, and it governs the whole analysis.
Dashboard — A page of tabs and widgets built over one or more datasets. See Dashboards.
Report — A document written from a project’s results, that can be presented as slides. Under construction. See Reports.
Widget — An element of a dashboard: an indicator, a figure, a table, an analysis, some code.
Plugin — A packaged, configurable, reusable analysis that becomes a widget. See Plugins.
Sharing
Export — A ZIP archive holding an entity’s tree of text files. Readable, comparable, re-importable. See Import and export.
Versioning — Tracking changes in a git repository: who changed what, when, and the ability to go back. See Git versioning.
Pointer — What a workspace keeps of an item linked to its own repository: its metadata and the repository address, not its content.
Mark for versioning — The gesture that explicitly includes a data file in an export or a repository. Without it, data stays with you.
Community catalog — The directory of published entities, installable in one click. Not to be confused with the data catalog, which documents a warehouse’s variables. See Community catalog.
AI and automation
Agent — An assistant you ask in plain language to act on a project. It acts with your rights, never more. See Agents.
MCP server — The service that gives an agent the tools to act on Linkr: search a concept, create a cohort, add a widget. See MCP server.
API key — A personal key, created in User settings › API keys, that lets an external client act as you without your password. It is shown only once.
Skill — A procedure written in plain language, which the assistant reads and applies. See Skills.
Background job — In server mode, a script run apart from the console, whose progress you follow while you keep working. See IDE.
Deploying
Client-only (browser mode) — Everything runs in your browser, with nothing to install. The data does not leave your machine, but there are no accounts, no sharing and no git. See Deployment modes.
Full-stack (server mode) — An installation with a backend: accounts and roles, shared storage, server-side code execution, git versioning, agents.
Data directory — On a server, the single folder where everything is kept. It is what you back up. See Backup and restore.
Secret key — The LINKR_SECRET_KEY string, which signs logins. Changing it signs everyone out, and loses nothing else.
Encryption key — The key that seals stored passwords and tokens: LINKR_ENCRYPTION_KEY, or failing that the secret.key file in the data directory. Losing it makes those secrets unreadable. See Configuration.
Going further
- Core concepts — the notions developed, rather than defined.
- Resources — the articles on OMOP, terminologies and how data is organised.
- Keyboard shortcuts — the other reference page.