Telemetry¶

What Dango and its bundled tools send home, and how to turn each one off.

Overview¶

Dango bundles four separate tools, each with its own telemetry: Dango itself, dbt-core, dlt, and Metabase. Telemetry is opt-in for Dango and disclosed-and-controllable for all four — this page documents exactly what each one sends, where it goes, and how to turn it off.

This page is about the product's telemetry — what Dango and the tools it runs send over the network while you use it. It's a different topic from the documentation site's privacy policy, which covers this docs website (hosting, cookies, fonts) and collects nothing product-related at all.

What's collected, per provider¶

The "what it sends" wording above is the same text shown in dango telemetry status and the /settings/telemetry web page — one source of truth, not three descriptions that can drift apart.

Dango's own ping fires twice: a one-time install ping (sent once, the first time you accept the consent prompt) and a weekly heartbeat while dango start is running. The install ping alone can't tell a one-time dango init apart from an install still in active use weeks later — the heartbeat is what makes "active install" measurable.

How to opt out¶

CLI — all providers at once:

dango telemetry off --all

CLI — a single provider:

dango telemetry off --provider dango
dango telemetry off --provider dbt
dango telemetry off --provider dlt
dango telemetry off --provider metabase

Check current state anytime with:

dango telemetry status

Web UI: the same four toggles are available at /settings/telemetry (admin-only).

One-time consent prompt: the first time you run dango init, Dango asks a yes/no question before sending anything:

Help improve Dango by sending anonymous usage data (no source names, credentials, or data — just install count)?

Your answer is stored in ~/.dango/telemetry.json and never asked again. Answering "no" (or hitting Ctrl-D / closed stdin) means nothing is ever sent — Dango's install ping only fires after an explicit "yes."

Environment variables: DO_NOT_TRACK=1 and DANGO_TELEMETRY=0 (or any other falsy value: f, false, n, no) both suppress every Dango ping, regardless of what's stored in ~/.dango/telemetry.json — confirmed in is_telemetry_enabled(). telemetry: false in ~/.dango/config.yml does the same. dbt-core additionally honors DBT_SEND_ANONYMOUS_USAGE_STATS=false and DO_NOT_TRACK=1 directly (dbt's own opt-out, independent of Dango's).

CI is always excluded. Dango detects common CI environment variables (CI, GITHUB_ACTIONS, GITLAB_CI, JENKINS_URL, BUILDKITE, CIRCLECI, TRAVIS, CODEBUILD_BUILD_ID) and never sends a ping from CI, regardless of any stored consent — CI runs are the single biggest source of inflated install counts, so they're excluded unconditionally rather than relying on the same opt-out logic as everything else.

What's NOT collected¶

Dango's own ping payload is limited to exactly what the table above lists: an anonymous install UUID, the Dango version, the OS name, the Python version, and configured source type names (e.g. "postgres", "stripe" — the category, not your actual source names). Specifically, it never includes:

• No source names, connection strings, or credentials
• No table names, row counts, schema, or query text
• No data values of any kind
• No personal information or email addresses

Telemetry is also best-effort and fails silently: the network call runs on a background thread with a 2-second timeout, and any failure — DNS, timeout, TLS, a bad response — is swallowed without retry. Telemetry can never block or break a CLI command.

Identity is machine-level (~/.dango/telemetry.json), not project-scoped — one person running Dango against five different client projects on one laptop counts as one install, not five.

The other three providers' payloads are scoped the same way — see the table above for exactly what each one sends. None of the four sends query results, credentials, or data values.

Why it exists¶

Dango is a small, largely unfunded open-source project. An anonymous install count — nothing more specific than "someone somewhere ran dango init" — is the only signal the maintainers have that anyone is actually using it. There's no other analytics, no user tracking, no product usage funnel. If you'd rather not contribute even that, every mechanism above turns it off completely.

Full network egress¶

Telemetry is one category of outbound network traffic; Dango and its bundled tools also make other network calls that aren't telemetry (dbt package installs, DuckDB extension loading, Metabase's Docker image pull, etc.) and aren't user-controllable because they're required for the software to function. The complete, verified list of every host Dango's stack contacts — telemetry and functional — lives in network-egress.yml in the main repo.

Next Steps¶

• Audit Logging — security event logging, a separate feature from telemetry
• Privacy Policy — the documentation website's privacy policy (a different topic from this page)