Stratix concepts — the mental model behind models, evaluations, judges, scorers, traces, and ECU.
Read these first. The mental model makes the rest of Stratix easy.
Mental model
How everything fits together.
Read
Models and benchmarks
The building blocks of the catalog.
Evaluations
What an evaluation actually is.
Traces and spans
The unit of "what did my AI do."
Judges
LLM-as-a-judge for subjective dimensions, and GEPA.
Scorers
Deterministic graders for objective dimensions.
Evaluation spaces
Saved evaluation configs.
Public vs Premium
What each surface owns.
ECU credits
How compute is billed.
BYOK models
Bring your own LLM endpoint.
Multi-tenancy
Orgs, multi-org, isolation.
Async vs sync workflow
When work is queued vs immediate.
Agentic evaluation
Pre-deploy gates for multi-step agents.
Continuous evaluation
Score live production traffic.
Workflow — the 6-stage spine that ties concepts together
Tutorials — concepts in action
Glossary — every term defined
Last updated 1 month ago
Was this helpful?
