File connectors
Parquet
- name: events
type: parquet
path: "./data/events.parquet" # single file
# path: "./data/events/*.parquet" # glob
# path: "s3://bucket/events/*.parquet" # S3 via DuckDB's httpfs extension
CSV
- name: users
type: csv
path: "./data/users.csv"
read_csv_auto handles header detection, delimiter inference, and type
inference. For anything non-trivial, convert to Parquet first — it's
faster, smaller, and carries schema.
S3 / HTTP
Parquet sources with an S3 or HTTPS path: use DuckDB's httpfs
extension. Set credentials via the usual AWS env vars (AWS_ACCESS_KEY_ID,
AWS_SECRET_ACCESS_KEY, AWS_REGION) or the DuckDB secrets mechanism.