Spreadsight turns messy spreadsheets into beautiful visualizations and actionable insights using AI. We built the full SaaS — auth with OTP, Polar billing wired through BetterAuth, S3 storage with presigned URLs, OpenAI-powered AI pipelines, dashboards, and a realtime sidecar for streaming AI progress to the UI.
Most teams have data in spreadsheets and questions they want answered. Most teams don't have an analyst on hand or the patience for pivot-table acrobatics. The gap between 'we have the data' and 'we have the answer' is where Spreadsight lives.
Upload a sheet. The AI infers the schema, surfaces the patterns, and gives you visualizations and insights you can share. No coding required, no copy-paste into another SaaS.
Column inference, type detection, semantic understanding — all done by the AI pipeline. The user uploads a file; the platform decides what's interesting and why.
Polar wired through BetterAuth so subscription state lives in the session, not a checkpoint the user hits later. Cleaner UX, simpler code.
A realtime sidecar service pushes AI progress to the dashboard as the pipeline runs. Users see the visualization assemble live instead of staring at a spinner.
Spreadsheets upload via presigned S3 URLs. The platform infers column types and semantic roles before any AI runs.
The AI SDK orchestrates structured generation against the inferred schema — visualizations, insights, summaries. Streaming responses flow into the realtime sidecar.
Subscription state lives in the user session. The paywall is implicit — the moment a feature requires it, the session knows.
Heavy AI work runs in a TSX-watched worker process. The UI reads from the realtime channel; the worker writes results.
Spreadsheet uploads land via presigned S3 URLs. Column types and semantic roles are inferred before any AI runs, so the pipeline starts with a structured view of the data.
An OpenAI-powered chat interface answers questions against the user's sheet, citing the cells it used. Thread history persists per file so the analysis is resumable.
A separate realtime service handles WebSocket fan-out for AI progress. The Next.js app reads from it; users see visualizations and threads update live.
Auth, billing, storage, AI pipelines, realtime — all wired through a single session layer. Subscription state lives in the session, not a separate paywall gate.
AI is the hard part on a build like this — Spreadsight proves we can ship it in production, not as a demo.
Senior team. Milestone-gated phases. You own the code at handoff. AI is the hard part — we've shipped it in production.
Book a 15-min demoFirst call is 15 minutes. We'll tell you whether the AI surface area fits a 2-week wedge or a 6-week flagship build.