Databow

databow

databow is a command-line tool for querying databases via ADBC (Arrow Database Connectivity). It exposes an interactive SQL shell as well as a non-interactive mode that is well-suited for use by agents and scripts.

Prerequisites

Install databow

databow must be installed. Check with databow --help.

Common install methods:

uv tool install databow
# or
pipx install databow
# or
cargo install databow

Install ADBC driver

An ADBC driver for the target database must be installed. The recommended way is via dbc, which installs drivers in a location databow discovers automatically:

dbc install duckdb
dbc install postgresql
dbc install snowflake
# ...etc.

If a query fails with a driver-not-found error, install the appropriate driver with dbc install before retrying.

Connecting to a database

There are two ways to specify connection details. Use whichever is more appropriate for the task.

Method 1: CLI flags

Specify the driver and connection options on the command line:

databow --driver duckdb --uri warehouse.db
databow --driver postgresql --uri postgresql://user:pwd@localhost:5432/db
databow --driver mysql --uri 'root:pwd@tcp(localhost:3306)/sys'

Available flags:

• --driver — ADBC driver name (e.g. duckdb, postgresql, mysql, sqlite, clickhouse, snowflake, bigquery, flightsql, mssql, redshift, databricks, trino).
• --uri — Database URI (driver-specific format).
• --username — Database username.
• --password — Database password.
• --option key=value — Driver-specific option. May be repeated.

Method 2: Connection profiles

ADBC connection profiles are TOML files that store driver and connection options, e.g. ~/.config/adbc/profiles/warehouse.toml:

profile_version = 1
driver = "clickhouse"

[Options]
uri = "http://localhost:8123"
username = "user"
password = "pass"

Use the profile by name (if it lives in a profile search location) or by path:

databow --profile warehouse
databow --profile /path/to/warehouse.toml

Prefer profiles when the same connection will be used repeatedly, or when credentials should not appear on the command line.

Common workflows

Quick ad-hoc query against in-memory DuckDB

databow --driver duckdb --query "SELECT 1 + 1 AS two"

Query a PostgreSQL database and export to CSV

databow \
  --driver postgresql \
  --uri postgresql://user:pwd@localhost:5432/app \
  --query "SELECT id, email, created_at FROM users WHERE created_at > now() - interval '7 days'" \
  --output recent_users.csv

Run a SQL file using a stored profile and write CSV

databow --profile warehouse --file reports/weekly.sql --output weekly.csv

Produce a Markdown table for inclusion in a report

databow --driver duckdb --mode ascii-markdown \
  --query "SELECT * FROM 'https://blobs.duckdb.org/data/penguins.csv'" \
  > penguins.md

Troubleshooting

• Driver not found / failed to load. Install the driver with dbc install (e.g. dbc install duckdb). Verify the driver name passed to --driver matches.
• Authentication errors. Double-check --uri, --username, --password, and any driver-specific --option values. For databases that require certificates, tokens, or extra parameters, see the upstream databases reference.
• Profile not found. Either pass an explicit path with --profile /path/to/profile.toml, or place the profile in an ADBC profile search location.
• Garbled or wrapped output. Use --output result.csv (or another file format) instead of stdout, or switch --mode to nothing / ascii-no-borders for plain rows.
• Hanging command. If databow was invoked without --query, --file, or stdin input, it has dropped into the interactive REPL. Re-invoke with one of those flags for non-interactive use.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: columnar-tech
• Source: columnar-tech/skills
• License: Apache-2.0
• Homepage: https://console.columnar.tech/agent-skills

Install and usage instructions live in the source repository linked above.