---
title: Getting started | DeepTable
description: Turn a messy spreadsheet into a clean SQLite database.
---

In this tutorial, you’ll upload a spreadsheet, convert it into a structured sheet, and download it as a SQLite database and individual CSV files.

## Prerequisites

- **Python 3.9+** or **Node.js 20+**
- A DeepTable API key (get one at [deeptable.com](https://www.deeptable.com))

## 1. Install the SDK

- [Python](#tab-panel-0)
- [TypeScript](#tab-panel-1)

Terminal window

```
pip install deeptable
```

Terminal window

```
npm install @deeptable/deeptable
```

## 2. Set your API key

Terminal window

```
export DEEPTABLE_API_KEY="your-api-key"
```

## 3. Download an example spreadsheet

Terminal window

```
curl -O https://docs.deeptable.com/acme_subscriptions.xlsx
```

## 4. Run this script

- [Python](#tab-panel-2)
- [TypeScript](#tab-panel-3)

Create a file called `main.py`:

```
import asyncio
from pathlib import Path


from deeptable import AsyncDeepTable




async def main():
    client = AsyncDeepTable()


    # Upload your spreadsheet
    file = await client.files.upload(
        file=Path("acme_subscriptions.xlsx")
    )


    # Create a structured sheet (this triggers the conversion)
    structured_sheet = await client.structured_sheets.create(file_id=file.id)


    # Wait for processing to complete
    while structured_sheet.status in ("queued", "in_progress"):
        await asyncio.sleep(10)
        structured_sheet = await client.structured_sheets.retrieve(structured_sheet.id)


    # Download as a SQLite database
    sqlite = await client.structured_sheets.download(structured_sheet.id, format="sqlite")
    await sqlite.write_to_file("acme_subscriptions.sqlite")




if __name__ == "__main__":
    asyncio.run(main())
```

Run it:

Terminal window

```
python main.py
```

Create a file called `main.ts`:

```
import DeepTable from "@deeptable/deeptable";
import * as fs from "fs";


const client = new DeepTable();


async function main() {
  // Upload your spreadsheet
  const file = await client.files.upload({
    file: fs.createReadStream("acme_subscriptions.xlsx"),
  });


  // Create a structured sheet (this triggers the conversion)
  let structuredSheet = await client.structuredSheets.create({
    file_id: file.id,
  });


  // Wait for processing to complete
  while (structuredSheet.status === "queued" || structuredSheet.status === "in_progress") {
    await new Promise((resolve) => setTimeout(resolve, 10000));
    structuredSheet = await client.structuredSheets.retrieve(structuredSheet.id);
  }


  if (structuredSheet.status === "failed") {
    throw new Error("Processing failed");
  }


  // Download as a SQLite database
  const response = await client.structuredSheets.download(structuredSheet.id, {
    format: "sqlite",
  });
  const buffer = Buffer.from(await response.arrayBuffer());
  fs.writeFileSync("acme_subscriptions.sqlite", buffer);
}


main();
```

Run it:

Terminal window

```
npx tsx main.ts
```

## What just happened?

You uploaded a spreadsheet, and DeepTable:

1. **Analyzed** the structure - finding tables, headers, and relationships
2. **Converted** the messy spreadsheet data into clean, normalized tables
3. **Returned** a SQLite database you or an LLM can query directly

## Download as CSV or Parquet

Need individual tables instead of a database? You can download each table separately:

- [Python](#tab-panel-4)
- [TypeScript](#tab-panel-5)

```
async for table in client.structured_sheets.tables.list(structured_sheet.id):
    csv = await client.structured_sheets.tables.download(
        structured_sheet_id=structured_sheet.id,
        table_id=table.id,
        format="csv",  # or "parquet"
    )
    await csv.write_to_file(f"{table.name}.csv")
```

```
for await (const table of client.structuredSheets.tables.list(structuredSheet.id)) {
  const response = await client.structuredSheets.tables.download(
    structuredSheet.id,
    table.id,
    { format: "csv" }  // or "parquet"
  );
  const buffer = Buffer.from(await response.arrayBuffer());
  fs.writeFileSync(`${table.name}.csv`, buffer);
}
```

## Next steps

- Explore the [API Reference](/api/index.md) to see all available endpoints