Limitations¶
Dango is honest about what it doesn't do yet. This page lists the real, current constraints — not vague caveats — along with whether each one is planned, and if so, roughly when.
Essentially nobody outside the maintainer has run this in production yet. Dango is early. Treat that as the headline limitation, not a footnote.
Known limitations¶
| Limitation | Status |
|---|---|
| Single-writer concurrency — Marimo cannot do concurrent read-write with the rest of the platform today | Planned. Fixed by the Quack migration (DuckDB's concurrent read/write storage format), targeted for when DuckDB 2.0 stabilizes (~September 2026). Waiting on DuckDB upstream, not on Dango. |
| Cloud deploy is DigitalOcean-only for managed, automated provisioning | Not planned. Bring-your-own-server (BYOS) works with any provider today via manual setup. Automated wizards for AWS/GCP/Hetzner are only revisited on user demand from those providers — none yet. |
| 34 data sources, vs. Airbyte's 300+ | Not scheduled. Sources are added when a specific user asks for one, not on a fixed roadmap or schedule. |
| No SSO / SAML — admin login, 2FA, and API keys only | Not planned. Revisited on enterprise customer signal — none yet. |
| No streaming, CDC, or reverse ETL | Not planned. Out of scope for what Dango is; revisited only on user demand. |
| Chrome-tested only — Safari and Firefox are untested, not merely unverified | Not scheduled. Sits in the demand-driven backlog, pulled in on user signal rather than a committed date. |
| No encryption at rest on the server — the DuckDB file is unencrypted | Planned, no date. A real roadmap item, independent of the Quack migration. Not a typical small-team need today, so it's further out than the "not planned" items above. |
Deploying model changes is blind — no dango diff or dev --remote yet | Planned. A core workflow gap, estimated at 3–4 weeks of work. |
| Scale ceiling is stated as "tested to X," not estimated | Not scheduled. Large-dataset performance testing hasn't been run past what's actually been exercised in practice — this isn't a number we're willing to invent. |
| Single project per server | Not scheduled. Multi-project-per-droplet sits in the demand-driven backlog. |
| BYOS deployments without a configured backup destination store backups locally only — if the server is lost, so are the backups | Partially addressed, full fix planned. Deploying without a backup destination now shows an explicit warning at deploy time. Full S3-compatible backup support for BYOS depends on two roadmap items — cloud storage configuration, then the BYOS backup destination itself — neither shipped yet. |
What this page doesn't cover¶
Category-level positioning (Dango isn't a petabyte-scale platform, a streaming system, or an ML environment) lives on the marketing site. This page is the specific, technical version — the things you'd actually hit using Dango, not the category it's not competing in.