Skip to main content
Docs/Docs
DocsBusiness Data Ingest
Pro

Business Data Ingest

Drop in a CSV, Excel, or PDF and GenMB extracts the shape, infers a schema, and proposes three ready-to-deploy app candidates. Skip the blank-page problem when you already have data.

Business Data Ingest requires a Pro or Business plan.

Supported Formats

CSV

First-class. Header row drives column names; values drive type inference.

Excel (.xlsx, .xls)

First sheet by default; the wizard lets you pick another if the file has multiple.

PDF

Via Google Document AI. Requires a form parser or OCR processor provisioned in your GCP project under the us location. Best for invoices, forms, structured tables.

The Ingest Flow

1

Drop the file

From the dashboard, choose "Start from data" and drag in your file (or paste a Google Drive link for Excel).
2

Review the extracted shape

The wizard parses the file and shows: detected columns / fields, inferred types, and 5-10 sample rows so you can sanity-check before generating.
3

Adjust the schema

Rename columns, change types (string → number, string → date), mark fields required or not. Edits flow directly into the generated database schema.
4

Pick an app proposal

Three candidate app shapes appear. Pick one and codegen runs with the schema and seed rows pre-wired.
Seed rows are loaded into the new app's PostgreSQL tables on first deploy, so the app shows real-looking data immediately rather than an empty state.

How Schema Is Inferred

Types are guessed from the actual values in each column:

  • Integers and floats for numeric columns.
  • Date / Timestamp for ISO-like strings or recognizable date formats.
  • Boolean for true/false, yes/no, 1/0 columns.
  • String for free text.
  • UUID when values match the UUID format.
If the wizard guesses wrong, override in the schema-edit step. Type changes are applied before provisioning, so the database is correct from day one.

App Proposals

Based on the data shape, GenMB proposes three different angles on the same data so you can pick the one closest to your need:

  • Public-facing: catalog, listings, or storefront powered by the data.
  • Internal admin: CRUD interface for your team to manage records.
  • Analytics dashboard: aggregated charts and KPIs over the same rows, often paired with Data Insights.

You can also start with one proposal and add the others later via chat ("also generate an admin view").

PDF via Document AI

PDF ingest uses Google Document AI. Two processor types are useful:

  • Form Parser: best for structured forms with named fields (invoices, applications, intake forms).
  • OCR processor: best for tables in scans where you mostly want text extracted in reading order.
Provision the processor in your GCP project under the us location. The platform caches processor lookups per type so repeated PDF ingests are fast.

Limits

  • File size: typical CSVs up to ~10 MB; larger files should be chunked.
  • PDF page count is bounded by your Document AI processor quota.
  • Sample rows shown in the wizard are capped (first 5-10) for performance; the full file is used to generate the schema and to seed.
  • Pro or Business plan required.

For ongoing data sync (Stripe, Shopify, Notion) rather than a one-shot ingest, see Data Connectors. For bulk-loading more rows into an existing app, see CSV Import.

FAQs

What document formats are supported?▾
CSV and Excel out of the box. PDF via Document AI (form parser or OCR processor) when configured in your GCP project. JSON is on the roadmap.
What does GenMB do with the parsed data?▾
It infers a schema from the columns or extracted entities, then proposes three app shapes you can spin up with one click: a public-facing site, an internal admin tool, or an analytics dashboard, depending on the data shape.
How is this different from CSV Import?▾
Business Data Ingest is the wizard entry point that takes a raw document and proposes whole apps. CSV Import is the runtime SDK (window.genmb.import) you call from inside an app to ingest more CSV rows into a table that already exists.
Does the source file get stored?▾
The parsed shape (schema + sample rows) is what powers the proposals; the original file is held only as long as needed to extract it. The generated app gets its own copy seeded into its database.
Can I refine the proposed schema before generating?▾
Yes. The wizard shows the inferred schema and the sample rows before code generation. Edit column names, types, or required-ness, then proceed.

Ready to build?

Create your first app for free, no credit card required.