Now the data scientists want all of it, and each dataset is a separate request.

Connect your data, query with Python and Arrow Flight, in five minutes

1. Install and start Provisa

Download the installer for your machine from the download section and run it. The first launch finishes a one-time setup with no internet needed. Then start the services and open the UI.

provisa start   # every service, on your machine
provisa open    # the UI, in your browser

2. Connect the sources

In the UI, open Sources and add Postgres, Elasticsearch and Kafka. Provisa reads each source's catalog; pick the tables to register. Nothing is copied: queries go to the data where it lives.

3. Install the client

In the notebook's kernel:

%pip install "provisa-client[pandas]"

4. Query into a DataFrame

Set PROVISA_URL and PROVISA_TOKEN in the shell before starting Jupyter, so no credential is saved in the notebook. Then load the cell magics and query:

%load_ext provisa_client.magics

%%sql
SELECT e.service, COUNT(*) AS errors, AVG(o.total) AS order_value
FROM logs e
JOIN orders o ON o.trace_id = e.trace_id
WHERE e.level = 'error'
GROUP BY e.service

5. Pull large results over Flight

For millions of rows, read over Arrow Flight into pyarrow, pandas, Polars or DuckDB:

from provisa_client import ProvisaClient
import os

client = ProvisaClient(os.environ["PROVISA_URL"], token=os.environ["PROVISA_TOKEN"])
table = client.flight("SELECT * FROM events WHERE day >= '2026-10-01'")

What stays true

• The notebook reads what the role reads, masked as the role's results are.
• To train without personal data, query the synthetic copy of a table instead.
• A query against one source goes straight to its driver.