v0.12.0

Minor Changes

• f001a3d: Make acrm_value writable with the obvious four-column INSERT (issue #51). The schema previously required attribute_type (already known from acrm_attribute) and active_from (mechanical bookkeeping), so the natural query failed with a validation error and new developers hit the wall on their first direct write.

  • acrm_value.attribute_type is removed. The type lives on acrm_attribute — join when you need it (acrm-query skill has the canonical pattern). The two internal read sites that pulled attribute_type from acrm_value (the dedupe flow's loadActiveValues / loadInboundRefs) now JOIN to acrm_attribute.

  • acrm_value.active_from is now Lix-defaulted to lix_timestamp(). Writers don't have to pass it. id was already defaulted to lix_uuid_v7().

  The naive insert from the issue now works:

  INSERT INTO acrm_value (object_slug, record_id, attribute_slug, value_json)
  VALUES ('people', 'person_1', 'name', '{"full_name":"Ada Lovelace"}');

  normalized_key / ref_object / ref_record_id stay as nullable indexed columns on acrm_value — direct-SQL writers still populate them for unique-keyed attrs and record-references (documented in the acrm-query skill).