Concepts/Concept
ADPS Agent Systems Blue Book · Engineering Concept
One-Diagram Unit of Work: Commit One Verifiable Modeling Increment
Commit one diagram, one resolved scope, one write set, and one receipt at a time.
Start with “finish the system model”
That goal may include requirements, use cases, block definitions, internal blocks, and state diagrams. It has no stable single-run completion condition, and an early deviation can propagate into later diagrams.
Definition
A one-diagram unit of work groups one diagram type, one resolved mount point, one write, and one acceptance receipt into a transaction boundary.
one user operation = one diagram type + one scope_ref + one WriteSet + one receipt
The boundary defines what may change, which invariants are checked, what the receipt counts, and what a rollback must restore. Larger goals proceed through multiple accepted units.
Boundary
This unit does not guarantee cross-diagram consistency. An upper-level plan still manages dependencies, versions, and acceptance order. If the host lacks atomic transactions, the adapter needs before-state snapshots, compensation steps, and an explicit partial-failure state.
Provenance
- Initial research source: the AI4MBSE personal research project by Liangding Yuan; the name is an ADPS synthesis.
- Lineage: Unit of Work, Command, and transaction boundaries.
- ADPS contribution: combines diagram type, engineering scope, candidate structure, receipt, and rollback into one agent write-back boundary.
- Definition status: candidate concept.
Initial source: AI4MBSE personal research project; authored by Liangding Yuan.