AI planned, deterministically computed
Turn data into dashboards in seconds.
Drop in a spreadsheet or connect an API. DashVault designs the charts and computes every number from every row.
DashVault turns a plain English request into a chart specification, then computes the chart from every row of the dataset. The model returns the plan; the engine returns the numbers.
The model sees
A small sample and an exact statistics digest. Never the full dataset, never your credentials.
The engine computes
Chart specs run through aggregate.ts over every row, in your browser or on the server.
The key stays put
One server-side environment variable. Nothing prefixed NEXT_PUBLIC, nothing in the bundle.
01/How it works
Three moves, no spreadsheet gymnastics.
01
Bring data
Drop a CSV or an Excel workbook, or point DashVault at a JSON API with auth, pagination and a data path. Saved connections keep credentials server-side and masked in every list.
02
Describe the outcome
Ask in plain English, or hand over a screenshot of the dashboard you already have. Claude replies with chart specs: what to group, what to measure, how to aggregate. Specs, never values.
› points by region, monthly
{ "groupBy": "region", "aggregation": "sum" }
03
Keep it living
Dashboards self-heal broken specs, filter globally across every chart, export to CSV and PDF, refresh on a schedule and travel by gated share link.
02/The engine
Don't trust it. Test it.
This block imports the production aggregation module and runs it over a 48-row sample, right here on this page. Change the spec; the same code that computes every DashVault chart recomputes.
Group by
Measure
Aggregation
Result
48 rows → 4 groups
4 groups computed. Total 37,115.
Same code path as production: the spec above runs through the aggregation engine over every row, and what you see is its return value. Claude writes specs like this one; it never writes the numbers.
03/For points programs
A loyalty ledger, fully accounted for.
Connect a partnership program once. DashVault pulls organisations, profiles, credits, debits and website activity with full pagination, then keeps six standing reports current.
Program Overview
Awards, redemptions and running balance, month over month.
Liability Forecast
Rolls the unspent balance forward and prices the breakage.
Rewards by User
Who earns what, ranked, filtered and exportable.
Awarded by User MoM
Every award as a user-by-month pivot table.
Redeemed by User MoM
The same pivot, for what flows back out.
Engagement
Sessions, active users and activity, day by day.
Flagship report
The forecast your finance team asks about
The Liability Forecast estimates monthly award volume and the redemption rate against the available balance, rolls the unspent balance forward, and reports outstanding liability plus present and projected breakage on a directly comparable basis. Scheduled insight emails carry it to your inbox.
04/Sharing
Share the report, not the spreadsheet.
One link, one report
A /r/ link shows that report alone, with your saved filters as the viewer’s starting point. No navigation, no wandering.
Gated by email
Allowlists decide who gets in and magic links do the signing in. Flip a share public only when it should be.
Never quietly stale
Shares backed by a live connection re-fetch and re-aggregate server-side once their snapshot ages past 24 hours.
CSV that matches the screen
Exports download exactly what the filters show. No hidden rows, no surprise columns.
data sources pulled per connected program
standing reports on every project
max age of a live share snapshot
numbers invented by the model
Ready when your data is
Put your numbers on the record.
Open the vault, drop in a dataset and ask for the dashboard you actually want.