Seeding and data quality

Seeding

Venues are seeded from OpenStreetMap:

• /api/bars/seed — the admin bootstrap route (legacy name), which upserts OSM-derived places directly into places.places.
• The platform Fundi worker — city-by-city seeding at scale across the continent.

Every seeded place gets an ownerEntityId: a tier-0 organization is auto-created in entity.entities for each place, so ownership edges exist from day one and the claim flow always has a target.

Junk-name policy

Some OSM records carry names that are not names — ".", "8318", stray punctuation. The rule:

• A name containing no letters in any script is junk. It is seeded with status: 'flagged' and isActive: false, queued for admin review, and never published live.
• A name containing letters in any real script passes — Arabic, Shona, Ndebele, Portuguese, and every other script are valid.

Note

Non-Latin names are correct data, not corruption. Arabic names in Egypt, Morocco, and Algeria and Portuguese names in Mozambique are exactly what the graph should hold. Never “clean” a name because of its script.

Category taxonomy

The taxonomy lives in places.categories and maps each categorySlug to one or more schema.org @type values via schemaOrgType. Examples:

categorySlug	schema.org @type(s)
restaurants	Restaurant, FoodEstablishment
national-parks	Park, TouristAttraction
tech-companies	Organization
education	EducationalOrganization
landmarks	LandmarksOrHistoricalBuildings

Places reference the taxonomy via primaryCategoryId and secondaryCategoryIds — categories are never duplicated onto place documents.