Configuration

Settings come from CLI flags (bound into Viper) and environment variables for sensitive fields. The environment prefix is P2PG. Nested keys use underscores instead of dots (for example db.password → P2PG_DB_PASSWORD).

Environment variables

These are required; the process exits with an error if any are empty:

Variable	Maps to	Description
P2PG_DB_PASSWORD	db.password	PostgreSQL password
P2PG_DB_USER	db.user	PostgreSQL user
P2PG_S3_ACCESS_KEY	s3.access.key	S3 access key
P2PG_S3_SECRET_ACCESS_KEY	s3.secret.access.key	S3 secret key

No other configuration is read exclusively from the environment in the current code path; non-secret options are set via flags below.

CLI flags (dataloader)

Database

Flag	Default	Description
--db-host	localhost	PostgreSQL host
--db-port	5432	PostgreSQL port
--db-name	postgres	Database name
--db-pool-size	1	Connection pool size; also caps how many Parquet files are processed in parallel

S3-compatible storage

Flag	Default	Description
--s3-endpoint	localhost:9000	Endpoint as host:port (no scheme)
--s3-region	none	Region string passed to the client
--s3-secure	false	Use TLS when talking to the object store

Data loader

Flag	Default	Required	Description
--schema	(empty)	Yes	Target PostgreSQL schema
--table	(empty)	Yes	Target table name
--bucket	(empty)	Yes	Bucket containing Parquet objects
--path	/	No	Key prefix inside the bucket; if it does not end with /, a trailing slash is appended internally
--batch-size	10000	No	Maximum rows per read/upload batch (must be > 0)
--table-create	false	No	If set, create the table with inferred columns when it does not already exist
--table-truncate	false	No	If set and the table exists, truncate it before loading

Validation rules

• data-loader.batch-size must be greater than zero.
• data-loader.path must not be empty after configuration (the CLI default / satisfies this).

If no Parquet files are found under the prefix, the command logs a warning and exits successfully without loading data.