Unit type: Phase
Epic: lattice-platform-development (lattice-platform-agentic-development-v0.2.md)
Unit ID: phase-2
Status: Rolling-wave placeholder — see phase-2-status.md
Sketch: phase-2-sketch.md
Depends on: Phase 1 exit gate
Full VP-level expansion scheduled at: P1.11.3 (Phase 1’s own close-out slice), per the epic’s stated rolling-wave methodology (Part 0.5)
Data gets in, questions get answered, lineage is provable. Milestones M4 (ingestion), M5 (query + lineage), M6 (content pipeline).
Authoritative slice-level detail (lighter than Phase 0/1 by the epic’s own design, expanded further at P1.11.3): epic Part 6. The persistence-compiler integration across P2.1/P2.3/P2.4 is already fully specified there (see the Part 6 preamble) and is not repeated here.
This plan intentionally does not contain a full test-case enumeration or a finished module-by-module README/doc table the way Phase 0/1’s plans do — producing one now would contradict the epic’s explicit reasoning for deferring Phase 2 detail until Phase 1’s store-SPI learnings land. What follows is scoped to what is already concretely decided.
docs/architecture (known now)| Document | Change | Producing slice |
|---|---|---|
docs/architecture/rdf-sparql-patterns-guide.md |
No change expected — Phase 2 consumes it, does not extend it, unless P2.1.4a’s implementation surfaces a guide gap the way Slice 2 of rdf-sparql-patterns-phase did |
P2.1.4a |
ontology/persistence/docs/platform-vocabulary-alignment.md |
Consumed, not produced, here (produced in Phase 0 per P0.3.9) | — |
| Anything P1.11.3 names as a new doc obligation for Phase 2 | Deferred | P1.11.3 |
| New subproject | Track | Note |
|---|---|---|
workers/mapping_plan_compiler |
C-05 | Python, design-time only, per P2.1 |
platform/plan-ir |
shared IR boundary | C-04/C-05 boundary |
platform/reasoning-validation |
C-13 | P2.2 |
platform/ingestion-gateway |
C-04 | |
platform/query-plane |
C-12 | |
platform/lineage |
C-19 | |
platform/metering |
C-16 | |
platform/push-gateway |
C-18 | |
platform/feedback-router |
C-15 | |
workers/content_pipeline |
C-06 | |
workers/agent_orchestrator |
C-14 | Build before C-06 per the epic’s explicit ordering note |
clients/client-java, clients/client-python |
Generated SDKs | P2.10.6 |
Full README content (beyond “one exists”) is written when each module’s first slice lands, per copilot-instructions.
solution-design-specification.md delta plan (known now)| SDS section | Change | Note |
|---|---|---|
| §1 Technical Capability Catalogue | Add ingestion, query/decision, lineage, metering, push, feedback-router capabilities | Full rows written when P1.11.3 expands this phase |
| §3 Data Architecture Summary | Add INGEST_STAGING graph family and the staging-never-queried rule |
P2.3.4 |
| §4.1 Component Inventory | Add every module in §3 above | Per introducing slice |
| §7 Robustness Design | Add the write-back loop cap and poison-stimulus policy once those land (Phase 3, not this phase — noted here only so the eventual Phase 2 close-out slice does not claim them) | N/A — cross-phase note |
A full delta table, matching Phase 0/1’s depth, is produced at P1.11.3 alongside the slice expansion.
Standard per epic Part 12, applied once slices exist to gate.
Depends on Phase 1 exit gate. Blocks Phase 3 (C-07/C-09 need C-04’s ingestion path and C-12/C-19 for query/lineage).