Visual Engineering Research

An open index of evidence, theory, and practice

research-document

M. First Vertical Slice

Research Area
General Research
Status
draft
Version
0.1
Priority
medium
Updated
2026-07-22
Source
docs/vnext/13-vertical-slice.md

M. First Vertical Slice

Per §22 and the SDE greenfield workflow: the smallest implementation that proves the architecture, not a feature-complete publisher.

M.1 Chosen corpus

visual-engineering, project composition-science.

Why this one, from measurement:

  • 23 published documents — small enough to reason about end to end
  • contains the only fully-resolving typed relationship in the authored corpus: 4 artifacts linked by source_rep: RP-COMP-005
  • spans four artifact types: research-execution-package, experiment-report (EX-COMP-011, EX-COMP-012), research-framework (DF-COMP-002), theory-registry (TH-COMP-005)
  • its documents originate frontier records, exercising Originates
  • IDs are declared, so /a/{id} is exercised — and some sibling documents lack ids, exercising /s/{path-key}

This gives two research object types to navigate between (§22.9) with real, non-synthetic data.

M.2 Scope

§ Requirement Slice implementation
22.1 discover a real corpus glob discovery over content/projects/composition-science/**
22.2 parse in F# front matter + body sections; snake/camel; unknown keys preserved
22.3 typed domain model Artifact, Section, ReferenceValue, StatusReading (F)
22.4 validate duplicate id (blocking); dangling ref, missing id, orphan (warnings)
22.5 resolve relationships SourceRep, RelatedDocument (all 5 value kinds), Originates; backlinks derived
22.6 machine-readable indexes manifest, artifacts, artifact/{key}, edges, edges/{key}, findings
22.7 static baseline page F#-generated HTML per artifact + project overview; readable with JS disabled
22.8 interactive Limen view engine (F#→WASM) + thin kernel; real declared boundary
22.9 navigate ≥2 object types experiment-report → its REP → back; REP → frontier records
22.10 source provenance every artifact page links to its exact source path
22.11 back/forward pushState/popState through the kernel
22.12 build + publish GitHub Actions → Pages, 404.html fallback
22.13 automated tests L's fixtures + a composition-science golden test

M.3 Explicitly deferred

  • other seven projects (config change only, no new architecture)
  • search (slice proves retrieval via indexes; search is additive)
  • Git-history timeline (§19) — see M.5
  • frontier view beyond the Originates link
  • collection/facet pages — the current ones keep working, untouched
  • retiring research-catalog.json v1.1 — shim remains

M.4 Exit criteria — measurable

  1. All 23 composition-science documents parse; zero unknown keys dropped.
  2. The 4 source_rep edges resolve — the current engine derives 0.
  3. Zero fabricated values: no 2026-07-22, no "General Research", no "unknown" author anywhere in the output.
  4. Round trip: experiment report → REP → back, with correct browser history.
  5. Two consecutive builds byte-identical.
  6. limen verify --strict passes with a non-empty boundary.
  7. Bounded retrieval: full context for RP-COMP-005 in ≤ 2 fetches, ≤ 20 KB.
  8. Every legacy URL for these 23 documents emits or redirects.
  9. Measured search-index and payload sizes recorded, to validate K.4.

M.5 Not blocked later

Per §19, Git history is not required now but must not be designed out. Provenance already carries an optional Commit, and artifacts are keyed independently of content, so a later pass can attach history without re-modelling. Slice 1 leaves the field present and unpopulated.