Archive v0.28 — Typed Research Facets & Policy-Aware Workflows

[WangEn]

Implements v0.28 as explicit-scope, evidence-first research discovery and policy interpretation.

• live public GET /v1/archive/research/facets with all-page counts of sealed completed non-calibration Runs matching every current filter
• nine typed facets: Provider, Model, Test, E-level (E4+/below/missing), region, account tier, effective service tier, service assurance, native-cost availability
• SQL aggregation over full matching scope; top 25 per facet with truncated metadata and nullable/unknown historical contexts
• public /research shows observed facet counts and filter links without implying ranking or uniqueness
• GET /v1/archive/research/collections/:id/assessment?policyVersion= reuses the v0.23 policy evaluator against the frozen Run IDs, original replication, and only prior same-context calibration
• immutable research collection detail adds explicit policy version selector and Run-level status/reasons/calibration/replication annotations, without a collective matched-set claim
• database/URL input guards; PostgreSQL integration counts across pages, scoped facets, empty populations and policy-version assessment; API tests for public access and malformed inputs
• docs, ADR 0024, v0.28 Roadmap implemented; v0.29 multi-page captures deferred

No historical evidence or saved manifest mutation. No heterogenous-Test leaderboard or composite score.