Add three open questions (OQ-17, OQ-18, OQ-19) covering attribute query
strategy, CRUD generation approach, and storage-operations bridge placement.
Create ADR-033 recording the v1 decision: JSON path queries for attributes
with hand-written CRUD for static tables.
Expand forward-look.md with Repository Layer Strategy section analyzing
three approaches (JSON path, native columns via dbtype, hybrid) and their
implications for the metagraph pattern. Add drizzle-graphql and dbtype
from-dbtype comparison showing neither handles dynamic schema-as-data.
Update overview.md with dbtype/ujsx in the dependency diagram, expanded
ecosystem context in the bridging pattern section, and new open questions.
Align open-questions.md: resolve OQ-17 and OQ-18 for v1 (ADR-033), add
OQ-19 as open, update summary counts and ADR impact table.
Graph type definitions as TypeBox Modules — the core architecture evolution
for @alkdev/storage. The SchemaBuilder is removed (no existing consumers),
replaced by direct TypeModule construction with Metagraph.Import() for
base attribute composition and Type.Composite() for node/edge type
specialization.
Key additions:
- metagraph-module.md: Module pattern, edge constraints as named entries,
SchemaBuilder equivalence, DB bridge contracts (moduleToDbSchema return
type, validateNode/validateEdge signatures), 10 design decisions (DD1-DD10)
- forward-look.md: pointer abstraction (ujsx ValuePointer analogy, JPATH
Module), dbtype table rendering relationship, ujsx as universal IR pipeline
Critical corrections from architecture review:
- Type.Composite uses IntersectEvaluated (intersection, not Object.assign
override) — overlapping keys with subtype relationships resolve correctly
- Type.Ref inside Type.Composite within a Module is verified working
- BaseNode/BaseEdge use Metagraph.Import() for same-package Modules (Option B),
not local re-declaration (no circular dep within same package)
- Edge constraints use Type.String() for node type name arrays (not Type.Ref) —
constraints contain names, not schemas
