Data model

Source of truth

• Cloud Spanner is the authoritative datastore for normalized PIMS and Bubble data.

Primary domains

Handler writes data for a wide range of entities, including:

• Organizations, clinics, households, patients
• Appointments, appointment types/statuses
• Team members, products, invoices
• Bubble entities (subscriptions, benefits, payment plans, etc.)

Pipeline status

• Pipeline runs update status timestamps in Spanner to support incremental sync.
• Status updates land in the Pipelines table keyed by organization_id and name (pipeline name), with last_modified used by Stream for scheduling.

Batch write behavior

• Spanner writes are chunked by SpannerBatchWriteService to honor the 80,000 mutation limit.
• SpannerLoader computes index-aware row costs per UpsertStep.

Data ownership

• Handler owns transformation and normalization logic, but not the source data.