navu-catalog 4.14.2

package/.claude/settings.local.json

@@ -0,0 +1,24 @@
+{
+  "permissions": {
+    "allow": [
+      "Bash(node_modules/.bin/tsx probe-scope.ts)",
+      "Bash(node_modules/.bin/tsx probe-mint.ts)",
+      "Bash(grep -rn \"extractScopeValues\\\\|\\\\.scope =\\\\|scope:\\\\|DiscoverCatalogArguments\\\\|resolveSearchScope\\\\|find-products-tool\\\\|buildParameterSchema\" src --include-dir=.)",
+      "Bash(time npm run build)",
+      "Bash(curl -sk --max-time 5 https://127.0.0.1:9200)",
+      "Read(//Users/kduffie/projects/navu-apis/**)",
+      "Read(//Users/kduffie/projects/**)",
+      "Bash(npx tsc *)",
+      "Bash(echo \"tsc exit: $?\")",
+      "Bash(npx vitest *)",
+      "Bash(npm run *)",
+      "Bash(echo \"lint exit: $?\")",
+      "Bash(echo \"lint: $?\")",
+      "Bash(env)",
+      "Bash(echo \"tsc: $?\")"
+    ],
+    "additionalDirectories": [
+      "/Users/kduffie/projects/navu-apis"
+    ]
+  }
+}

package/.cursor/rules/harvest-strategy.mdc

@@ -0,0 +1,26 @@
+---
+description: Catalog harvest, product builder, manifest, and index guidelines — read before implementing or extending site parsers or harvest pipelines
+globs: src/site-catalogs/**/*,src/site-catalogs/**/*,src/reusable/**/*
+alwaysApply: false
+---
+
+# Catalog pipeline guidelines
+
+When implementing or modifying a **site harvester**, **product parser**, **product builder**, **CatalogManifest**, or **staging behavior**, read and follow:
+
+**[docs/catalog-pipeline-guidelines.md](../../docs/catalog-pipeline-guidelines.md)** · **[docs/catalog-search-architecture.md](../../docs/catalog-search-architecture.md)**
+
+Site-specific modules: `src/site-catalogs/<site>/`. Shared upstream clients: `src/reusable/shopify/`.
+
+Key reminders:
+
+- Target shape: **`CatalogHarvestProduct`** (named `fields` + `variants[]`) — not EAV `CatalogCleanProduct`.
+- One **OpenSearch document per product**; SKUs live in nested `variants[]`.
+- **`CatalogManifest`** drives mapping, filters, and discovery (`CatalogProperty` flags: grain, facet, scope, volatile).
+- Three AI tools per catalog: `{base}-discovery`, `{base}-query`, `{base}-get-details`. Query filters: `{ token, op?, value, source, stated_span?, implied_span? }`.
+- Stream staging via `context.staging.upsertItem` when using `xtras.catalog.*`; persist harvest rows to `catalog.harvestProducts`.
+- After harvest + manifest: run **`--build-index-documents`** → persist derived index JSON to **`catalog.openSearchDocuments`** (includes **`narrative`** for BM25; same as OpenSearch except no `narrative_embedding`); then **`--rebuild-catalog-index`** when OpenSearch is available.
+- Inspect real upstream payloads; use native types in `fields` (numbers, booleans, string arrays).
+- **One narrative per product** for embeddings (1536d, `text-embedding-3-small`); no price/stock in narrative.
+- Bump **manifest / processing version** when harvest or manifest semantics change; rebuild index after.
+- Phase B profiling + Phase D audit: see **catalog-site-review** skill.