Skip to content

Upgrading from Groundworkers 0.6

Groundworkers 0.7 uses the oa-configurator 2.0 stack and supports a vocabulary database separate from the primary CDM database.

Configuration changes

  • Rename the former database schema_name setting to cdm_schema for a CDM database. Use vocab_schema when vocabulary has a distinct schema. Generic databases, including vector-store databases, still use schema_name.
  • SQLite has no schemas. Remove schema settings from SQLite database entries; setup ignores those fields and tells you when they are unavailable.
  • A PostgreSQL vector store colocated with the CDM uses the dedicated groundworkers schema. Its database is a generic entry on the CDM connection:
[databases.groundworkers_vectors]
kind = "generic"
connection = "cdm_main"
schema_name = "groundworkers"
  • Add databases with the 2.0 CLI syntax, for example:
omop-config databases add cdm clinical --connection cdm_main --cdm-schema public
omop-config databases add generic vectors --connection cdm_main --schema-name groundworkers
  • Startup now fails fast if either the primary or vocabulary endpoint cannot be reached. The console prints the first scrubbed cause and a database recovery hint.
  • A pre-3.0 embedding store reports LEGACY_REGISTRY. The migration script is tracked in oa-configurator issue #40; once it ships, Groundworkers will provide the omop-emb migration command.
  • To split the vocabulary, set vocab_connection on the CDM database entry. Groundworkers 0.7 routes vocabulary reads to that connection, including setup diagnostics and graph searches.

Plugin changes

See Upgrading plugins to Groundworkers 0.7 for the full API checklist. Plugins should require groundworkers>=0.7,<0.8, pass cdm_session_factory for OMOP graph/vocabulary access, and treat embedding_backend_factory() as a read-only EmbeddingStoreReader.