Systemiq is agent-based. Different agents handle different parts of the stack, so configuration should not be treated as one global settings object.

Configuration is attached to the agents and components responsible for a specific task. The resulting model and memory behavior emerges from how those configured agents work together.

• Generic agents handle common tasks such as data cleaning, augmentation, ML inference, LLM reasoning, and automation.
• Custom agents implement client-, workflow-, or domain-specific logic where reusable defaults are not sufficient.
• Each agent is configured according to the part of the stack it owns and the context it is expected to derive.

Configuration layers

Configuration is layered so that one part of the platform can be adjusted without redefining everything else.

• Client-wide configuration defines defaults and shared behavior within the authenticated client boundary.
• System-level configuration defines how a business system is modeled and interpreted.
• Tool-level configuration defines how a specific tool processes or exposes context.
• Agent-level configuration defines how a specific agent performs its task, such as cleaning, ML, LLM reasoning, or automation.
• Indicator-level configuration defines how outputs are named, structured, and queried downstream.
• Access control determines which users and services can query or modify platform state.

Operational consequence

• Changing one agent does not require redefining the whole platform.
• Different parts of the stack can evolve at different speeds.
• The resulting model stays interpretable because ownership remains attached to a concrete scope.

Why this matters

• Agents can be tuned independently without changing the whole platform.
• Generic and custom agents can coexist in the same stack.
• The configuration model stays modular and interpretable even as the system grows.

Example configuration flow

A typical setup might model one business system with multiple tools and agents working together. For example, one tool may ingest operational data, one agent may clean or augment it, another may apply ML or LLM-based reasoning, and another may trigger automation based on the result.

• The system defines the business domain and upper context boundary being modeled.
• Each tool defines a specific subsystem or context step within that system.
• Each agent defines how that step behaves, for example cleaning, inference, reasoning, or automation.
• Indicators define how outputs are named, structured, and consumed downstream.

Implementation model

In most deployments, system modeling and configuration are defined together with Systemiq experts. This keeps agents, tools, and indicators aligned with the real business context rather than treated as a generic plug-and-play setup.