| Term | Meaning |
|---|---|
| Project | Authorized collection of an application’s simulations, Tools, datasets, and history |
| Environment | Selected execution/reporting scope within a project |
| Recipe | The target, scenarios, repetitions, configurations, and optional surroundings |
| Scenario | A named situation with explicit task input and optional sequence or assessment |
| Configuration | Named parameters supplied to your target, such as model settings |
| Batch | One admitted expansion of a recipe |
| Run | One scenario/configuration/repetition execution |
| Target | Your real command, HTTP application, native callback, or prepared isolated runner |
| Tool | A controlled synthetic dependency with operations and state |
| Build | An immutable Tool behavior and schema definition |
| Dataset | A versioned set of starting records for a compatible build |
| Instance | A running or retained synthetic Tool copy |
| Connection | A scoped, expiring grant for a Tool interface |
| Epoch | The state generation used to reject stale Tool access after restore |
| Evidence | Actual retained observations, operation outcomes, and captured files |
| Exact check | A hard comparison or requirement against recorded evidence |
| Rubric | Authored criteria and anchors for assessing meaning or behavior |
| Assessment | Recorded exact-check results and optional quality scores |
| CI gate | Optional policy decision, separate from execution and assessment |
| Operation key | Stable request identity used to reconcile the same intended action |
Glossary
The small set of terms used across Firedrill’s CLI, SDKs, MCP, and reports.
A scenario, a run, and an assessment are different records. A completed run does not mean every expectation held.