Data tool integration guides

Back to README data tools.

Start a local server or install a release. SQL needs Redis with Query Engine (Search); 4× QPF is recommended where supported. SQL uses the Redis ADBC driver's data model, and runs in autocommit mode without transactions. Named Arrow datasets are separate from SQL tables.

Local examples assume an unauthenticated endpoint without TLS:

Flight / ADBC: grpc://localhost:8815
JDBC: jdbc:arrow-flight-sql://localhost:8815/?useEncryption=false

Use an address reachable from the client. Inside a Docker Desktop container, host.docker.internal reaches a Flight server on the host; localhost refers to the container itself. For authenticated or TLS deployments, follow the connection options and server configuration.

Run and save results for the 27 inventoried Flight SQL routes with the repeatable integration harness.

Guides

Integration Guide
Python / PyArrow SQL and native datasets
Go Flight SQL and native Flight
Java JDBC application
DBeaver Driver setup and manual tests
Other JDBC applications Custom driver configuration
pandas SQL reads and parameter binding
Polars SQL reads and batch iteration
DuckDB Local SQL over downloaded Arrow
R ADBC driver manager
Apache Superset Version-specific connection guidance
Tableau JDBC setup and read preview
Metabase Community plugin and Actions
Power BI Custom connector development path
Redis Insight Inspect the backing Redis database
NumPy Install, read and write
scikit-learn Install, read and write
SQLAlchemy Install, read and write
Dagster Install, read and write
Jupyter / JupyterLab Install, read and write
Hugging Face Datasets Install, read and write
Streamlit Install, read and write
SciPy Install, read and write
XGBoost Install, read and write
PyTorch Install, read and write
TensorFlow Install, read and write
Keras Install, read and write
Hugging Face Transformers Install, read and write
Plotly Install, read and write
Matplotlib Install, read and write
Kubeflow Pipelines Install, read and write
Metaflow Install, read and write
Apache Spark Install, read and write
Apache Airflow Install, read and write
Quarto Install, read and write
RStudio Install, read and write

Every inventoried route has a guide above, including installation and a read/write workflow. For read-oriented or unavailable application connectors, the guide states the write limitation and provides an external ingestion path. A local model/artifact save is separate from a database write. The live recipe run records executed read/write scopes and remaining application limitations separately from the earlier historical sweep.

Optional sample SQL table

The pandas, Polars, DuckDB, Go, Java, JDBC and BI guides use flight_tool_sample. Run these commands from the repository root against a disposable backend. The table name must be unused; mode="create" refuses to replace an existing table.

python -m venv .venv-tools
. .venv-tools/bin/activate
python -m pip install -r examples/python/requirements-clients.txt
python - <<'PY'
import adbc_driver_flightsql.dbapi as sql
import pyarrow as pa

with sql.connect("grpc://localhost:8815", autocommit=True) as conn, conn.cursor() as cur:
    cur.adbc_ingest("flight_tool_sample", pa.table({
        "id": [1, 2, 3],
        "label": ["one", None, "three"],
        "amount": [1.5, 2.5, 3.5],
    }), mode="create")
PY

Keep the table while following the guides. An ordered SELECT should return three rows, with a NULL label on row 2; SUM(amount) is 7.5. When done, drop only the fixture you created:

python - <<'PY'
import adbc_driver_flightsql.dbapi as sql
with sql.connect("grpc://localhost:8815", autocommit=True) as conn, conn.cursor() as cur:
    cur.execute("DROP TABLE flight_tool_sample")
    cur.fetch_arrow_table()
PY

These guides describe setup and expected checks. Historical passing results, versions and application limits remain in the compatibility matrix. Creating a fixture through Python does not establish write support in another tool.