@unified-product-graph/core 0.9.10 → 0.9.11

package/CHANGELOG.md

@@ -7,12 +7,27 @@ This project uses [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ---
 
+## [0.9.11] - 2026-06-10
+
+**Tooling DX (batch 6, part 2): status pre-flight + anti-pattern visibility.** Two discoverability fixes from the same real portfolio session, plus a self-trip fix.
+
+### Added
+- **`list_status_values(entity_type)`** (tools 118 to 119, on both the local and cloud servers): the valid `status` values a type can hold, as a focused pre-flight lookup so an author no longer learns the set only from a rejected write. Returns lifecycle phases as `{ status, label, terminal }` plus `initial_status` / `terminal_statuses`, or `lifecycle_free: true` with empty `values`. Sourced from `UPG_LIFECYCLES` (exactly what the write validator checks): the low-token sibling of `get_lifecycle`.
+- **Anti-pattern version tracking.** `UPGCuratedAntiPattern` gains an optional `since` field, and `get_spec_version` now reports `anti_patterns: { total, versioned }`. A consumer can see which validators are newer than the version a graph was authored under, so a spec upgrade no longer silently flips a clean graph invalid with no heads-up.
+
+### Changed
+- **A drafted hypothesis is auto-promoted to `active`** when it gains a `hypothesis_requires_experiment_plan` edge (via `create_edge` / `batch_create_edges`). Pairing a hypothesis with a plan is what activates it; this stops the documented structural-spine recipe (hypotheses paired with plans) from self-tripping the `untested-hypothesis-pile-up` anti-pattern. Best-effort: a promotion failure never fails the edge.
+
+No entity, domain, or edge-catalogue count change (tool count 118 to 119).
+
+---
+
 ## [0.9.10] - 2026-06-10
 
 **Tooling DX + a born-valid product (batch 6, part 1) + a release-test gate.** Polish from a real multi-graph portfolio session: parity and discoverability fixes, one model correction, and — the headline — the release train now runs the unit suites, after they were found to have been silently red across a release.
 
 ### Changed
-- **Product `properties.stage` is canonical.** The stale `UPG_PROPERTY_MIGRATIONS['0.2.13']` lift of a product node's `properties.stage` to top-level `status` is retired: it contradicted the UPG-654/661 convention the whole runtime uses, so a freshly-created product read as `property_drift` against its own validator. A new product is now born valid. Legacy values are still normalised on read by `coerceProductStage`. Decision: `2026-06-10-product-stage-properties-canonical.md`.
+- **Product `properties.stage` is canonical.** The stale `UPG_PROPERTY_MIGRATIONS['0.2.13']` lift of a product node's `properties.stage` to top-level `status` is retired: it contradicted the current convention the whole runtime uses (a product node's lifecycle stage lives in `properties.stage`), so a freshly-created product read as `property_drift` against its own validator. A new product is now born valid. Legacy values are still normalised on read by `coerceProductStage`. Decision: `2026-06-10-product-stage-properties-canonical.md`.
 - **`get_nodes` resolves cross-product ids.** A qualified `{product_id}/{node_id}` (the form `list_registry` / `export_edges` / cross-edges return) now reads that product's graph (read-only for non-active products), so a connective pass fetches node content across the portfolio without a `switch_product` sweep. Results carry a `product_id`.
 - **`create_edge` / `batch_create_edges` accept the `p_` product-header** where the in-graph product node is expected, resolving it to the `type:"product"` node (older graphs carry a distinct `n_…` product node).
 - **`did_you_mean` for a mistyped edge type** now falls back to the closest catalogue key by name (edit distance) when the endpoints don't resolve to a canonical pair — so a typo always gets a concrete suggestion.

package/dist/index.cjs

@@ -24093,6 +24093,7 @@ var UPG_ANTI_PATTERNS = [
   // RelationshipCheck (edge presence/absence) and are promoted here.
   {
     id: "insights-without-evidence",
+    since: "0.9.7",
     name: "Insights without evidence",
     description: "The graph has insight entities but none are backed by a primary-evidence link: no observation yields them (`observation_yields_insight`), no survey response evidences them (`survey_response_evidences_insight`), and no quote is attached (`insight_evidenced_by_quote`). An insight with no evidence behind it is an opinion wearing a research label.",
     structured_condition: {
@@ -24136,6 +24137,7 @@ var UPG_ANTI_PATTERNS = [
   },
   {
     id: "feature-requests-without-provenance",
+    since: "0.9.7",
     name: "Feature requests without provenance",
     description: "The graph has feature_request entities but none trace back to a source: no feedback program collects them (`feedback_program_collects_feature_request`), no customer feedback becomes one (`customer_feedback_becomes_feature_request`), and none originate from a behavioural segment (`feature_request_from_behavioral_segment`). A request with no provenance cannot be weighed against who asked or how many.",
     structured_condition: {
@@ -27196,7 +27198,7 @@ function isCanonical(text) {
 }
 
 // src/index.ts
-var UPG_VERSION = "0.9.10";
+var UPG_VERSION = "0.9.11";
 var UPG_FORMAT_VERSION = "0.4.0";
 var MARKDOWN_FORMAT_VERSION = "0.1";
 var UPG_TYPES = getTypes();