rdra-ish-buc-update

name	rdra-ish-buc-update
description	Update an existing BUC by adding or modifying use cases, business flows, requirements, screens, fields, API contracts, concepts, NFRs, events, entities, ADR links, and predicates while preserving staged refinement
license	MIT

Update an existing BUC

Given a description of what to add or change, modify the relevant .rdra files while keeping the model consistent.

Preserve the model's current abstraction level. If the BUC is still a skeleton, do not force columns, APIs, or lifecycle rules. Ask for the next missing information required by ../../docs/incremental-modeling.md and apply the smallest stage-appropriate diff. Keep the business-to-technical refinement order intact: do not introduce technical details until the business value, actors, use cases, and data touchpoints that justify them are present.

Step 1 — Read the existing BUC

Read buc/buc_<name>.rdra and the shared files it imports. Identify:

• Which use cases already exist
• Which entities are already referenced
• What screens and events are already declared
• Which flow/step, field, DTO, requirement/NFR, concept, and ADR elements already exist
• Which predicates are already defined
• Whether the model still uses the small layout (shared/actors.rdra, shared/biz.rdra, shared/entities.rdra) or has split shared files

Step 2 — Classify the change

Change type	Where to edit
New use case in existing BUC	buc/buc_<name>.rdra
New business flow order, branch, exception, or loop	buc/buc_<name>.rdra with flow, step, precedes, branches, excepts, repeats, covers
New requirement metadata	stable shared requirement file, then motivates(Requirement, Buc)
New screen for an existing UC	buc/buc_<name>.rdra
New screen field or input/output mapping	owning BUC/screen file; add field, contains(Screen, Field), maps_field
New API method/path or payload	owning API file; add API metadata, dto, request, response, error_response
New concept/domain term	shared conceptual file; add maps_to only when logical data mapping is known
New NFR/quality/constraint	stable shared file; scope with applies_to, qualifies, or constrains
New ADR impact	stable shared ADR file; connect with decides(Adr, Target)
New entity column	shared/entities.rdra
New entity entirely	shared/entities.rdra + add relate if needed
New event or state	shared/entities.rdra (if cross-BUC) or buc/buc_<name>.rdra
New actor	shared/actors.rdra
New system/API ownership	shared vocabulary for system, BUC/shared API file for api, then contains(System, Api)
New Business-BUC context	shared vocabulary for location / timing / medium, then

If the model still uses the small layout, keep using it unless the change makes a shared file hard to review. If shared files are already split, place new shared definitions near their owning area and mirror path/module names.

Also classify the abstraction transition:

Current state	Concern	Next useful update	Ask the user for
BUC exists but no actors	Biz intent/value	actor coverage	who performs or receives value from the BUC
Actors/use cases exist but no CRUD	Biz object touchpoints	data touchpoints	entities and CRUD intent per use case
CRUD exists but no screens/API	Tech interaction boundary	interaction boundary	screens, external interfaces, API endpoints, owning systems, access/media constraints
Entities have only id	Tech data design	entity structure	fields, keys, relationships, cardinality
Structured entities have lifecycle fields	Tech lifecycle design	lifecycle	states, events, use-case effects
Lifecycle reaches plausible patterns	Tech-enforced rules	local guardrails	invalid or mutually exclusive state facts via forbidden / exclusive
Local guardrails are stable	Tech-enforced rules	obligations and advanced rules	invariant, narrow required, comparison propositions, or cross-entity rules when needed

Step 3 — Apply the minimal diff

Add only what is needed. Do not remove or rename existing predicates unless the requirement explicitly says to delete functionality.

Adding business flow order:

flow <Flow> "<flow label>"
step <StepA> "<business step>"
step <StepB> "<business step>"

contains(Buc<Name>, <Flow>)
contains(<Flow>, <StepA>)
contains(<Flow>, <StepB>)
precedes(<StepA>, <StepB>)
covers(<StepB>, <ExistingUC>)

Use branches, excepts, and repeats only when the requirement explicitly introduces an alternative, exception, or loop.

Adding a use case:

usecase <NewUC> "<action description>"

contains(Buc<Name>, <NewUC>)
<crud>(<NewUC>, <Entity>)
displays(<NewUC>, <Screen>)

Adding a column to an existing entity:

// in shared/entities.rdra — append inside the entity block
  <col_name>: <Type>  @null   // or other annotations

Then add sets in the BUC file for every use case that writes this column.

Adding a new entity with a relationship:

// in shared/entities.rdra
entity <NewEntity> "<label>" {
  id:   Int @pk
  ...
}
relate(<Parent>, <NewEntity>, 1:N)

Then in the BUC file add CRUD predicates for the use cases that touch it.

Adding or changing API/system boundaries:

system <System> "<system label>"
api <Api> "<API label>" method PATCH path "/orders/{id}" idempotency idempotent mode sync auth bearer
dto <RequestDto> "<request label>" {
  id: Int
}

contains(<System>, <Api>)
invokes(<UC>, <Api>)
request(<Api>, <RequestDto>)
updates(<Api>, <Entity>)
sets(<UC>, <Entity>, column == value)

System entity sets are derived from API CRUD. Add owns(System, Entity) only when intentional ownership must be visible before complete API operations exist, and review the warning if explicit ownership disagrees with derived ownership.

Adding a screen field:

field <Field> "<field label>"
  access editable
  required true
  source actor

contains(<Screen>, <Field>)
maps_field(<Field>, <Entity>, "column_name")

Adding Business-BUC context:

timing <When> "<timing>"
location <Where> "<place or channel>"
medium <Medium> "<device or terminal>"

belongs(Buc<Name>, <Business>)
  .when(<When>)
  .where(<Where>)
  .by(<Medium>)

Adding access constraints:

permission <Permission> "<permission>"

has_permission(<Actor>, <Permission>)
requires_permission(<UC>, <Permission>)
requires_medium(<UC>, <Medium>)
requires_permission(<Api>, <Permission>)

Use requires_* on the use case for constraints that apply to the whole interaction, and on the API for constraints specific to that backend boundary. Screen constraint patterns are derived through displays and invokes; do not hand-write a separate screen predicate for them. Actor-side grant gaps are derived with rdra-ish csv src/ --kind actor-permission-audit; use rdra-ish csv src/ --kind permission-callables to confirm the permission maps to the intended use cases and APIs.

Adding non-functional or decision traceability:

nfr <Nfr> "<label>"
  metric "latency"
  target "p95 <= 300ms"
  slo "99.9%"

quality <Quality> "<quality label>"
qualifies(<Quality>, <System>)
applies_to(<Nfr>, <Api>)

adr <Adr> "<decision label>"
decides(<Adr>, <Api>)

Starting a BUC from an event:

triggers(<Event>, <TargetBuc>)

If the entry action is already known:

contains(<TargetBuc>, <EntryUC>)
triggers(<Event>, <EntryUC>)

Handling a cross-system relation:

coordinates(<UC>, <EntityA>, <EntityB>)
invokes(<UC>, <ApiForEntityA>)
invokes(<UC>, <ApiForEntityB>)

Use this only when <EntityA> and <EntityB> are related and belong to different derived system boundaries.

Adding cross-entity rules:

forbidden(<EntityA>, <EntityB>,
  (<EntityA>.<column>, <value>),
  <EntityB>.<column> > <EntityA>.<column>)

invariant(<EntityA>, <EntityB>)
  .along(<EntityA>, <EntityB>)
  .when(<EntityA>.<column> == <value>)
  .then(<EntityB>.<column> == <value>)

Use bare column names only when the rule has a single-entity scope. Multi-entity rules must qualify columns as Entity.column. Use .along(...) only when the rule is about instances linked by a declared relate path; current states reports such relation-scoped rules as CrossConstraintNotEvaluated instead of checking the global cross-product. TLC evaluates .along via *_owner when relate exists — use rdra-ish export --kind tla for that check.

Removing a use case:

1. Delete the usecase declaration
2. Delete contains(Buc, <UC>)
3. Delete all CRUD, displays, raises, sets predicates referencing it
4. If the UC was also referenced by triggers or performs elsewhere, update those too

Step 4 — Consistency checks after the change

• Every new usecase has a contains predicate
• Every new flow belongs to a BUC, every step belongs to a flow, and covered behavior exists.
• Every new usecase has at least one CRUD predicate
• Every new field belongs to a screen; actor-entered fields are mapped or have an explicit reason to remain external.
• API method/path metadata has request/response DTOs when contract review is in scope.
• Every new entity column that can change has a sets or participates in transitions
• No dangling imports (if you removed the last use of a symbol, remove its import)
• No manually added FK columns for a relate-covered relationship
• Every API used as a system boundary has contains(System, Api)
• Every explicit owns(System, Entity) is either satisfied by API CRUD or accepted as future ownership with diagnostics reviewed.
• Every new permission used by an actor or operation is declared once in shared vocabulary unless intentionally local
• Every requires_medium references a declared medium, and screen constraints are checked with rdra-ish csv src/ --kind screen-constraints
• Permission-to-callable mappings are checked with rdra-ish csv src/ --kind permission-callables
• Actor grants are checked with rdra-ish csv src/ --kind actor-permission-audit; review both missing and excess
• Every cross-system relate has a coordinates(UseCase, Entity, Entity) when a use case handles the consistency
• Every coordinates use case invokes APIs on both system sides
• Every cross-entity rule qualifies columns as Entity.column and names or implies the participating entities
• Every .along(...) cross-entity rule names a declared relate path and is treated as CrossConstraintNotEvaluated until linked-instance reachability is implemented
• The diff does not jump more than one abstraction level unless the user supplied the missing information explicitly
• New shared files follow path/module correspondence, e.g. with

Step 5 — Validate

rdra-ish check src/
rdra-ish lint src/ --format table

Fix every reported error before declaring the update done.