Get started
Concepts
Installation, workspaces, roles, contexts, agents, flows, and schedules.
On this page
This page introduces the vocabulary used in the platform and throughout the documentation.
Installation and organization
An installation is a complete Aptabit Platform stack on a server, with its configuration, its data, and its license identity. Each installation serves one organization, created at the first sign-in.
To separate environments or business units with full isolation, use separate installations. Several installations can run on the same server, each with its own Compose project, containers, and network.
Workspaces
The workspace is the unit of work inside the organization:
- Personal: every person has one.
- Shared: created by an administrator, with members added directly or through corporate directory groups.
Contexts, flows, schedules, and integrations are organized by workspace. A model appears in a workspace's chat only after it is enabled there. The workspace picker sits in the top bar.
Roles and modules
Access combines three layers:
- Role: root account (owner of the installation), administrator, or member.
- Per-person modules: Chat, Agents, Storage, Data, Flows, Schedules, and Integrations. New people start with no modules.
- Scope: workspace membership and access to each agent and model.
Contexts
A context is where the files and data the platform uses are kept:
- Knowledge: documents read, split into chunks, and indexed for hybrid search (semantic and lexical).
- Data: typed tables, filled from files or by hand, queried through read-only SQL.
- Personal: each person's fixed area with the files they attached in chat, visible only to them.
See Storage: knowledge and data.
Models and providers
An external provider is a model service reached through an API: OpenAI, Azure OpenAI, Anthropic, or any endpoint compatible with the OpenAI API. A local model runs on your server, in a runtime managed by the platform itself or in an inference server you already operate. See Models, runtimes, and accelerators.
Agents
An agent combines a base model, instructions, and generation parameters, with access restricted to workspaces and people. Knowledge comes from the context chosen in the conversation or from a flow built around the agent. See Agents.
Flows, contracts, and runs
A flow is an automation assembled from blocks in a visual builder. Its contract defines the exposure (internal, API, or chat) and the JSON input and output schemas. A flow moves from draft to published and runs only when published.
Each run produces a run record with source, status, input, output, and the trace of every block. See Flows and schedules.
Schedules
A schedule runs a published flow at a fixed interval (from 60 seconds), with a fixed JSON input, up to a maximum number of runs if one is set.
Integrations
An integration is a registered connection to an external system, with its resources and encrypted credentials. Flows use it through the integration block. See Integrations.
Logs and provenance
The unified log brings together every chat, agent, and flow interaction and the audit events. Provenance is what each AI answer keeps about its origin: the model that produced it and the sources that grounded it. See Governance and compliance.
License
The license is issued by Aptabit to your organization and validated online, through check-ins, or offline, through a file. See License and connectivity.