Skip to content

Reference & Data Model

Overview

Automation & AI comprises configuration data (who may do what with which means) and runtime data (what was executed when).

The entities of this chapter are organized in three layers:

  • Capabilities: Agent (agent), Agent skill (agent_skill), MCP server (agent_mcp_server) with MCP tools (agent_mcp_tool), LLM provider (llm_provider) with LLM models (llm_model). They describe which AI building blocks are available to the organization.
  • Automation: Workflow definition (workflow_definition) as the header of a rules workflow, Workflow deployment (workflow_deployment) as a stored version of its rules document and Trigger (workflow_trigger) as the start condition.
  • Runtime: Execution (workflow_execution with workflow_execution_step) for every run and Conversation (agent_conversation) for chats with Polaris or an agent.

Polaris itself is not an entity. The built-in assistant lives in code and has no row in agent. Skills and MCP servers are assigned to it via the flag „Use in Polaris“ or „In Polaris verwenden“.

llm_provider ──has many──▶ llm_model ─ ─ ─ (model ID as text) ─ ─ ▶ agent.model
│
agent_skill ◀──n:m (agent_skill_map)──────────────── agent ──1:1──▶ app_user (service account)
agent_mcp_server ◀──n:m (agent_mcp_server_map)───────┘
│
└──has many──▶ agent_mcp_tool (Allow / Ask / Deny)
workflow_definition ──has many──▶ workflow_deployment (v1, v2, …)
│ │
└──has many──▶ workflow_trigger ◀───┘ (exactly one per deployment version)
│
▼ starts
workflow_execution ──has many──▶ workflow_execution_step
▲ (child runs via parent_run_id)
agent_conversation ─────────────┘ (chat run, trigger „Chat“)
└──belongs to (optional)──▶ agent
Entity Purpose Where in the app
Agent (agent) Your own AI specialist with instructions, model, skills, servers Aligia AI › Agenten › Agents → Guides › Agents
Agent skill (agent_skill) Reusable instructions for Polaris and agents Aligia AI › Agenten › Agenten-Skills → Guides › Agent skills
MCP server (agent_mcp_server) Connection to an external system via Model Context Protocol Aligia AI › Agenten › MCP Servers → Guides › MCP servers
MCP tool (agent_mcp_tool) Single function of an MCP server with permission detail page of the MCP server
LLM provider (llm_provider) Your own contract with a language model provider Aligia AI › Agenten › LLM Provider → Guides › LLM provider
LLM model (llm_model) Concrete model of a provider with facts and prices detail page of the LLM provider
Workflow definition (workflow_definition) Rules workflow (rules document with trigger), written by Polaris No page of its own → Guides › Workflows
Workflow deployment (workflow_deployment) Immutable version of the rules document version label of an execution
Trigger (workflow_trigger) Start condition of a workflow (manual or data change) No page of its own → Guides › Triggers
Execution (workflow_execution) One run with steps, child runs and transcript Page Ausführungen (/executions) → Guides › Executions
Conversation (agent_conversation) Private chat with Polaris or an agent Aligia AI › Polaris → Guides › Polaris
  • An active agent owns a service account (app_user, origin “agent”), see Service accounts.
  • Permissions: creating and changing MCP servers, LLM providers and workflows, as well as activating agents, are reserved for owners and admins. Executions and conversations are private, see Permissions and Private data.
  • Organization-wide defaults (default models, platform models) live in the Organization settings.
  • Workflows read and write business data of all other chapters; database triggers react to their changes. The interplay with the engine’s automation is described in Chapter 12.
  • Delegated work on issues and tasks runs through conversations, see Issue and Task.