Documents do not map one-to-one to capacity
One source document may produce many searchable passages. A product may have several representations or relationships. Planning from the number of original files alone can hide the size of the retrieval model you intend to build.
Take a representative sample through the actual preparation process. Count the resulting rows, searchable points, and graph relationships rather than assuming a universal multiplier for every dataset.
Track the three dimensions separately
Polygres project capacity includes storage, AI Context, and graph dimensions. Storage covers database data and structures. AI Context capacity applies to collection points. Graph capacity weights active nodes and edges.
A project can be constrained by one dimension while using little of another. A text-heavy corpus and a densely connected entity graph may need very different capacity selections even if they began with a similar source record count.
Use the free-to-paid transition correctly
The free selection is a complete Nano configuration. Increasing a dimension promotes the project selection to the Basic minimums across all dimensions. It is not a separate paid add-on attached to an otherwise free project.
The current Basic base includes 2 GiB of storage and 200,000 points or units for AI Context and graph. Use the calculator to see the monthly estimate as you change the dimensions together.
Revisit the estimate when the model changes
Changes to chunking, graph density, or the selected source tables can alter capacity needs. Record the assumptions behind the estimate so you can revise them when the application evolves.
Treat the dashboard quote as the billing reference for the project. The public calculator is useful for planning, while account credits and first-period adjustments belong to the account-specific flow.
Explore the next step
Continue with the project capacity calculator. For the current setup and API contract, use the Polygres documentation.