Network Access

Visivo is built to work without an internet connection. Installing it is the only step that needs one: after that, visivo compile, visivo run, visivo serve, visivo test and visivo dist all work offline, against whatever sources you can reach.

This page lists the exceptions, so you know what to expect on a disconnected laptop or behind a proxy that only allows your own hosts.

Works offline

The CLI	compile, run, serve, test, dist, init
The viewer	Dashboards, charts, tables, inputs, the editor and its SQL autocomplete
Local sources	DuckDB, SQLite, CSV and Excel files on disk
Static builds	A visivo dist bundle serves from any static host with no outbound calls

Everything the browser needs is served from your own Visivo process. The SQL editor, the DuckDB engine the viewer runs queries in, and the hot-reload channel are all part of the install.

Needs the network

Map tiles

Map charts draw their background from a tile service — carto-positron by default, with mapbox, stadiamaps and arcgisonline available as theme styles. A world basemap is tens of gigabytes, so it is not something an install can carry.

Offline, a map chart still plots its data and shows "Map background unavailable" on the chart. Nothing fails silently, and the rest of the dashboard is unaffected.

To avoid the dependency entirely, use a chart type that does not need a basemap — choropleth and scattergeo draw their own geography from Plotly's built-in outlines.

The agent

visivo agent and the agent tab send your prompt and the project context to a model provider. There is no local model, so the agent needs a connection and an API key. Everything else in visivo serve works without it.

Screenshots the agent takes of your dashboards need a headless browser. A pip install 'visivo[screenshot]' downloads it once on first use; a packaged build cannot download it and says so, pointing at PLAYWRIGHT_BROWSERS_PATH.

Visivo Cloud

visivo deploy, visivo archive and visivo authorize talk to app.visivo.io. These are the commands whose whole purpose is to put a dashboard somewhere other people can see it, so a connection is inherent to what they do.

Your own data sources

A warehouse source — Snowflake, BigQuery, Postgres, MySQL, Redshift — is reached over the network by definition. Local file sources are not.

Excel uploads

Reading a .xlsx file uses DuckDB's excel extension, which DuckDB fetches from extensions.duckdb.org the first time it is needed and then caches in ~/.duckdb/extensions/. If it cannot be fetched, the upload says so and suggests CSV, which needs nothing.

Telemetry

Anonymous usage events go to Visivo unless you turn them off. They are fire-and-forget: nothing waits on them and nothing fails when they cannot be sent. See Telemetry to opt out with visivo telemetry off.