Schema Intelligence - Pidgeon Health

Flock reads your schema before it generates anything. It classifies each table as patient, encounter, clinical, financial, or reference, maps the foreign keys into a dependency graph, and infers what each column holds: MRN fields, coded values, date columns, free text. That graph is what keeps every generated row FK-safe.

Supported databases

Flock introspects three database engines. Schema inference, sample extraction, and seeding are implemented for each.

Database	`--provider` value
PostgreSQL	`postgres`
SQL Server	`sqlserver`
MySQL	`mysql`

MariaDB connects through the mysql provider. postgresql and mssql are accepted as aliases for postgres and sqlserver.

Connect

pidgeon flock connect \
  --provider postgres \
  --connection "Host=localhost;Port=5432;Database=ehr;Username=dev;Password=${DB_PASSWORD}"

What Flock reads

Connecting runs a full introspection pass against the database catalog:

Table classification. Each table is tagged patient, encounter, clinical, financial, or reference from its columns and naming.
Relationship mapping. Foreign keys become a dependency order, so an insert never precedes the row it references.
Column semantics. Flock identifies MRN fields, date columns, coded values, and free text from types and constraints.
Healthcare patterns. Common EHR naming conventions are recognized so the classification holds on real vendor schemas.

Run pidgeon flock schema --classify to print the classification, or pidgeon flock schema --visualize to emit the schema as a Mermaid ER diagram.

API

# Connect and introspect
curl -X POST http://localhost:5102/api/flock/connect \
  -H "Content-Type: application/json" \
  -d '{"provider": "postgres", "connectionString": "Host=localhost;Database=ehr;..."}'

# Retrieve the classified schema
curl http://localhost:5102/api/flock/schema