Glossary (manifests & experiments)

Terms you will see in YAML files and CLI output every day. Each user guide also explains them in context; this page is the single reference list.

Term	Meaning
Manifest	A YAML file that declares an agent (agent.yaml), a multi-agent team (mas.yaml), an experiment (experiment.yaml), an overlay, a flavour, etc.
MAS	Multi-agent system — several agents with a workflow (who talks to whom). Configured in a mas.yaml manifest.
Agent	Single LLM actor with tools, skills, and plugins. Configured in an agent.yaml manifest.
Overlay	A small YAML patch (kind: Overlay) merged on top of a manifest. CLI: -o path/to/overlay.yaml. Used to vary one knob (tools, memory, governance) without copying the whole file.
Flavour	Deployment preset: model endpoint, instrumentation, transport. Referenced from experiments or config.yaml (e.g. standard:openai).
Lab (definition)	A folder *.lab/ (under labs/) with experiment.yaml, datasets, overlays, and optional custom pipeline steps. Experiment surface — not a library. See labs-and-libraries.md.
Library	A folder with library.yaml. Reusable apps, tools, plugins, and code. name:path always uses a library name. May live inside a lab.
Package	A published wheel (mas-runtime, mas-ctl, mas-lab, …). See packages reference.
Component	One of the three lab sub-wheels under lab/components/ (mas-lab-core, mas-lab-bench, mas-lab-controller). Distinct from the internal mas.lab.* runtime modules — see ADR 0001.
Experiment	The experiment: block (usually in experiment.yaml): what to run, how many times, and which pipeline builds results.
Model slots	experiment.models (and lab.models): named LLM pins matching Agent/MAS spec.models[].id, plus judge. Scalar experiment.model = models.main. any inherits the next layer; local config.yaml defaults.model fills remaining any at engine time.
Scenario	One column in an experiment matrix — a named setup (id) and layered overlays (logic / control / infra). Declared under scenarios:.
Item	One dataset case (item: hooks, CLI --item). Formerly called test:.
Dataset	Input items (prompts, expected fields) the experiment iterates over. Referenced as dataset: in experiment.yaml.
Run	One execution of the agent/MAS for a given (scenario, dataset item, repeat index). Produces traces/events.jsonl.
n_runs	How many times to repeat each (scenario, item) for variance. Declared under run.n_runs.
Benchmark	mas-lab benchmark run — executes all runs, then runs level hooks.
Pipeline	Ordered pipeline steps that read run artifacts and write CSV/PNG under results/.
Level hooks	run.post / item.post / scenario.post / experiment post: in experiment.yaml. There is no pipeline: key on the experiment.
Pipeline step	One unit in a level hook (e.g. extract_trace_stats, plotnine). Declared with name, type, config.
events.jsonl	Machine-readable run log: one JSON object per line (model calls, tool calls, governance, routing).
Exchange log	Human-readable transcript on stderr during mas-ctl chat (--trace = summary + timestamps; --trace full for the payload dump). Persist with mas_ctl.trace in config.yaml. Not used for experiment scoring.
Trace cache	Store of completed run logs; identical inputs reuse the cache instead of calling the model again.
Observability	Settings that enable events.jsonl (manifest spec.observability, overlay, or --events).

Runtime internals: see the runtime/docs/ tree in the repository.

Hub: references/index.md · Run logs: cli/observability.md.