원클릭으로
adbc
Connect to and work with databases using Arrow Database Connectivity (ADBC). Use whenever the user needs to use a database.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
메뉴
Connect to and work with databases using Arrow Database Connectivity (ADBC). Use whenever the user needs to use a database.
Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.
Query external databases from DuckDB using the ADBC (Arrow Database Connectivity) community extension. Use when the user wants to run SQL against any database with an ADBC driver (BigQuery, ClickHouse, Databricks, PostgreSQL, MySQL, Redshift, SingleStore, Snowflake, SQL Server, SQLite, Trino, etc.) from within DuckDB, attach an external database as a DuckDB catalog, read remote data with `read_adbc`, execute arbitrary remote DDL/DML with `adbc_execute`, or manage ADBC connection profiles/drivers via `dbc`.
Query SQL databases with the databow CLI. Use when the user wants to run SQL queries against any database with an ADBC driver (BigQuery, ClickHouse, Databricks, DataFusion, DuckDB, Exasol, Microsoft SQL Server, MySQL, PostgreSQL, Redshift, SingleStore, Snowflake, Spark, SQLite, Trino, etc.), export query results to JSON/CSV/Arrow IPC files, or work with ADBC connection profiles.
Install ADBC (Arrow Database Connectivity) drivers with dbc. Use when the user wants to install database drivers and connect to databases.
SOC 직업 분류 기준
| name | adbc |
| description | Connect to and work with databases using Arrow Database Connectivity (ADBC). Use whenever the user needs to use a database. |
| license | Apache-2.0 |
| metadata | {"author":"Columnar","version":"0.2.2"} |
Use the dbc command line tool to find available drivers for a database. It's possible no driver may exist for the user's database of choice as not all databases have ADBC drivers.
If the user does not have dbc available, try to install it for them.
Prefer installing it with these commands, in order of preference, if the tool is available:
uv is available: uv tool install dbcpipx is available: pipx install dbcbrew is available: brew install columnar-tech/tap/dbcwinget is available: winget install dbcdbc search
Install a driver by running dbc install <DRIVER>. This is idempotent — if the driver is already installed, it is a no-op that simply reports the existing installation — so there is no need to check first before running it.
Prefer to install drivers using dbc over installing driver packages from PyPI or Conda Forge.
For anything beyond a one-shot dbc install <DRIVER> — reproducible dbc.toml / dbc sync workflows, version pinning, or any other dbc subcommand — invoke the dbc skill rather than guessing at commands. In particular, note that there is no dbc list command.
Refer to drivers by their dbc short name and avoid specifying drivers with absolute paths.
Example: After running dbc install sqlite, refer to that driver as sqlite. For example:
from adbc_driver_manager import dbapi
dbapi.connect(driver="sqlite", db_kwargs={"uri": "foo.db"})
Don't use dbc info to find where drivers are installed and find their absolute paths on disk.
See the resources below depending on which language or languages the user wants to use:
resources/languages/cpp.mdresources/languages/go.mdresources/languages/javascript.mdresources/languages/python.mdresources/languages/r.mdresources/languages/rust.mdPrefer any language the user has already said they're using or can use. Each of these examples use the "sqlite" driver, connect to an in-memory database, and load a "penguins.parquet" file. This is done just for explanatory purposes and the code should be adapted to the user's problem.
See the resources below depending on which database the user wants to use:
resources/drivers/duckdb.mdresources/drivers/flightsql.mdresources/drivers/mysql.mdresources/drivers/postgresql.mdresources/drivers/snowflake.mdresources/drivers/sqlite.mdConnection profiles are TOML files that store a driver name and connection options, referenced by a profile://profile_name URI. They decouple credentials and environment-specific config from application code.
Suggest connection profiles when the user:
All language bindings that wrap the C++ or Rust driver manager support connection profiles, including Python, Go, R, Java, GLib/Ruby, C++, and Rust. JavaScript (@apache-arrow/adbc-driver-manager) also supports connection profiles via the profileSearchPaths option.
See resources/connection-profiles.md for the TOML format, file locations, and environment variable substitution syntax. See the relevant language resource for the binding-specific API.