core_concepts

Scope: Closed vocabulary for core concepts.
Audience: Contributors and reviewers.
Guarantees: Definitions are canonical and enforced.
Non-Goals: Expanded domain theory.
Why: This doc exists to record its single responsibility for review.

Overview

Core concepts are defined as a closed vocabulary.
Read Docs Style before edits.
Architecture components are defined in Architecture.

Contracts

Definitions below are the only canonical wording.
All docs reference this page for vocabulary.
Terms are linked when used.
- agent: An agent is a stateful decision unit that selects a tool and emits actions.
- tool: A tool is a callable capability used by an agent within an execution unit.
- policy: A policy is a constraint set that governs agent decisions and tool usage.
- execution unit: An execution unit is a bounded step that applies a tool and records outputs.
- evaluation artifact: An evaluation artifact is a stored result from scoring or validation.
- protein: A protein is a constrained agent with explicit state and constraints.
- signal: A signal is a typed input that triggers a state transition.
- pathway: A pathway is a multi-agent system of protein agents.
- regulation: Regulation is bounded proposal tuning that never mutates state.
- cell: A cell is an execution environment for pathways.

Invariants

Canonical wording is enforced in Naming.
Aliases are not used in docs.
Vocabulary aligns with Getting Started.
failure semantics align with Invariants.

Failure Modes

Alias usage fails docs lint.
Unlinked terms fail docs lint.
Redefinition outside this page fails review.
recovery semantics align with Invariants.

Extension Points

Change requests follow Triage.
Updates propagate to Getting Started.
Updates propagate to Architecture.

Exit Criteria

This doc becomes obsolete when vocabulary is generated.
The replacement is Architecture.
Obsolete docs are removed.

Code refs: tests/unit/test_docs_contract.py.