FAQ

Today: PostgreSQL, Neo4j, and files via CSV / Excel.

Coming soon: MySQL, Elasticsearch, and JSON / XML.

If a system has no dedicated connector but can export, CSV / Excel takes an upload, a URL Chartizer fetches on a schedule, or a drop into an S3 bucket.

No. Run a local agent inside your own network. It holds the credentials locally, executes queries there, and returns only results. It makes outbound connections only, so there are no inbound firewall rules to open.

Never. Every connector is read-only. Connect with a role that can SELECT and nothing else — the Sources page has the exact grants.

Per dataset: live query at render, on a schedule from 15 minutes to daily, on file upload, or manually. Live queries are freshest but inherit your source’s latency, so large tables belong on a schedule. See refresh policies.

A source is one connected system plus its credentials — created once, reused everywhere. A dataset is a named view over part of that source: a table, a view, a query result, or a parsed file. Projects read datasets, never sources directly. See Core concepts.

Not yet. Combining datasets requires a shared mapped field, and field mapping is coming soon. Until then, reconcile upstream — a database view or consistent naming at the source — so one dataset already carries what you need.

Almost always one source, not many datasets. When credentials expire, every dataset beneath that source falls into Auth error. Fix the source once with Reconnect and they all resume — clear authentication problems before investigating anything else.

You give Chartizer a question; it plans an approach, then works through your connected data with its own read-only tools — catalogue search, field profiling, aggregate queries, joins on mapped fields, anomaly detection — and returns a written finding with the charts and the exact queries behind each claim. Each query is presented for your approval before it runs.

It also reports what it could not determine, and flags caveats such as a partial period or a sync gap. See Investigations.

No. Every connector is read-only and no tool writes, so an investigation can read, compute, and chart, but never modify a source or a dataset. It also runs as you — it cannot reach data you could not open yourself, and row-level rules still apply.

It works from metadata plus a small sample — field names, types, and up to 50 sampled values per field — not full tables. For Ask your data, your question and the generated query are processed; result rows are not retained.

Every query is shown to you and runs only once you accept it. The assistant writes queries; it never executes one on your data without a person approving it first.

Sampled values are real customer data. If that is unacceptable for a dataset, disable sampling on it. Full detail in Data handling.

Yes, per feature and team-wide under Settings → AI, or per dataset with Exclude from AI features. Members cannot re-enable what an Admin has disabled.

No. Every suggestion waits for a person to accept it. A wrong mapping silently corrupts joins, so nothing lands unreviewed — see Mapping suggestions.

No, and this is a common misunderstanding. An agent keeps credentials and query execution on your infrastructure. If sampling is enabled, the agent still returns sampled values to Chartizer. For a database that must stay fully private, run the agent and disable sampling.

Yes. Permissions are configurable: an admin can toggle individual capabilities on a built-in role, or create custom roles from Settings → Roles.

The one exception is Admin, the superuser — it always holds every permission and cannot be restricted, so a team can never configure itself out of access to its own billing and settings. See Roles and permissions.

Only its members, unless it is shared more widely. Team role and project role are separate: an Admin can manage every source and still not open a project they are not a member of. See Roles and permissions.

A share link is an invitation, not access. The first time someone opens it they must request access, and a team admin approves that individual before anything renders. Forwarding the link does not pass access along — the next person gets their own approval prompt.

The one setting that removes the gate is auto-approve domain, which lets anyone with an address at that domain straight in. Use it for your own domain only. Every approval is logged — see Sharing and embedding.

The source was connected with credentials tied to that person — a database login in their name, or an API key they created. When the account was disabled upstream, every dataset under it fell into Auth error.

Connect sources with a dedicated service account and a role that exists independently of any individual. Doing this from the start is the single best habit for avoiding surprise Authentication expired errors — see Roles and permissions.

Yes. Sign a token on your server with the viewer’s identity and attributes, then load the embed iframe. Row-level rules apply, so each of your users sees only their own data. Never let the signing secret reach the browser. See Sharing and embedding.

Active sources. Pausing one frees a slot while keeping its datasets, their last synced data, and every chart built on them. Limits by plan are in Team settings.

Deletion cascades: removing a source removes its datasets, which breaks charts reading them. Chartizer names every affected project before you confirm — read that list.

For files backing a dataset, replace rather than delete and re-upload; replacing keeps the dataset identity and its charts.

Timezone and week start, number and date formats, sensitive-field tagging, and defaulting new projects to Private. Timezone especially — changing it later re-buckets every date-grouped chart.