Skip to main content
Log in

Send feedback

Tell us what works and what could be better. We read every submission.

Send feedback

Tell us what works and what could be better. We read every submission.

Projects and data

Projects group datasets, visualizations, AI work, and related activity. Use a project when you want to visualize your own dataset or collaborate around a larger task.

Create a project

  1. Open Project Quick Start or Create New Project.

  2. Drop one CSV, TSV, JSON or Parquet dataset onto the large upload target, or click to choose it, then select Next.

  3. VizHub inspects up to ten records and suggests a project name and description using an LLM. This uses the creator's personal or organization's AI credits; the page discloses the sample sent to the provider. Wait on the generation screen, or choose Retry / Continue manually if generation fails or credits are unavailable.

  4. Review and edit the prefilled name and description on the existing project form. Set visibility, organization access where applicable, and optional extra files.

  5. Select Create Project. The dataset uploads through the existing project creation path, which starts Analyze Data and opens AI Edit Review. Analysis uses AI credits separately from metadata generation.

  6. The separate Continue with full automation after Analyze Data checkbox remains unchecked by default when visible. The project-level opt-in controls further idea generation and prototyping after you approve the first analysis.

To create a project without a dataset, choose Create without a dataset on the first screen; this opens the original manual form.

Organization projects can be created through the relevant organization workspace when your role permits it.

Upload datasets

The standard project uploader accepts CSV, TSV, JSON, and Parquet files, with a 10 MiB per-file limit. For the easiest first experience, use CSV with a header row and consistent field types. Text and Markdown files can be references but are not automatically analyzable datasets.

After uploading, inspect the dataset preview and field names. If a field appears as text instead of a number, review the source data before building a quantitative chart.

For large Parquet files, the preview may be bounded rather than showing the entire dataset; it is not a full export of every row.

Analyze Data and review

For a new project created with a dataset, Analyze Data starts automatically. You arrive directly in AI Edit Review; generated variations appear as they become available. Check each candidate against your actual data and approve the result you want to keep.

For a project created without a dataset, upload data first and then choose Analyze Data from Project Tools. In this manual workflow you review and explicitly submit its AI prompt.

If automatic analysis cannot start (for example, insufficient credits), the created project remains available with a link back to the AI workflow so you can retry deliberately. It does not automatically retry and spend more credits.

Depending on permissions and available features, you may continue with Generate Ideas and Prototype Ideas, then review their outputs.

Keep data and permissions aligned

A public visualization can expose data used to render it. Before making a visualization public, confirm that its dataset is appropriate for public sharing and that you have rights to publish it. A private project should not be treated as public simply because an individual chart has a share control.

AI activity can spend credits owned by you or your organization. A project may include other collaborators; only users with suitable roles can edit, approve, share, or change settings.

Related: AI Edit and review, Sharing and embeds, and Agents and conversations.