AWS S3 is Amazon's object storage service, storing over 500 trillion objects and handling 200 million requests per second. It's the foundation for data lakes, analytics pipelines, AI training datasets, and application storage at any scale.

The S3 Wrapper lets you query files stored in S3 directly from Postgres as foreign tables. Read CSV, JSON Lines, and Parquet files with plain SQL and join them against your application data. Works with any S3-compatible provider, including Supabase Storage, Wasabi, Cloudflare R2, and Backblaze B2. It is read-only and works with Supabase Vault for secure credential management.

Supported File Formats

Format	Notes
CSV	With or without header line. All columns must be defined in the foreign table as text.
JSON Lines	All columns must be defined in the foreign table as text.
Parquet	Native type mapping supported. Compressed files are loaded fully into memory, so keep file sizes small.

Supported Compression Algorithms

gzip, bzip2, xz, zlib

Required S3 Permissions

• s3:GetObject
• s3:ListBucket

Supported Data Types for Parquet

Postgres Type	Parquet Type
boolean	BOOLEAN
integer	INT32
bigint	INT64
real	FLOAT
double precision	DOUBLE
numeric	DECIMAL
text	BYTE_ARRAY / UTF8
date	DATE
timestamp	TIMESTAMP_MILLIS / TIMESTAMP_MICROS

Resources

• supabase/wrappers on GitHub