Skip to main content
The Data API runs SQL against your Catalog tables over HTTP, so any application or service can read your data in Nekt. You submit a query and get back links to download the results.
The Data API is available on all paid plans.
Your data never transits through Nekt servers. Results are written to your organization’s cloud storage and returned as presigned URLs that expire after 1 hour.

How it works

1

Create an API key

Go to your workspace settings and create an API key. Every request authenticates with it in the x-api-key header.
2

Submit your query

The Data API runs on AWS Athena (AWS) or Google BigQuery (Nekt Express and GCP), using standard SQL. Choose the output format: CSV (default) or Parquet.
3

Download the results

The response includes one or more download links for the result files.
The download links are only valid for 1 hour.

Run a query

Request body

Available mode_options depend on your cloud and the output format:

Response

Output formats

On Nekt Express and GCP, CSV doesn’t support complex types (ARRAY, STRUCT, JSON). Use Parquet for those columns.

Known limitations

Timestamp with time zone. Nekt stores dates and datetimes with timestamp precision, which can fail with:
Cast timestamp columns in your query:
Duplicate column names. Parquet doesn’t allow duplicate column names (Column name 'column_name' is duplicated). Alias columns so every name is unique:
Columns with only NULL values. Parquet can’t infer the type (Unable to infer schema for Parquet). Cast the column explicitly:
Complex nested types. Deeply nested structures (arrays of structs, maps with complex types) may not serialize correctly. Flatten them, or convert them to JSON strings:
Very large strings. String values over 2 GB can run out of memory. Filter or truncate large text fields, or use CSV.
Timestamps. BigQuery may format timestamps differently than Athena. Cast timestamp columns the same way:
High-precision decimals. Very high precision decimals may be rounded. If exact precision matters, use CSV or cast to string:
  • Special characters: CSV uses standard escaping. Check fields with commas, quotes, or line breaks after download.
  • Data types: every value is a string, so convert types when you import the file.
  • Decimal precision: very large decimals may lose precision as strings. Review them after import.
  • Timeout: queries that run longer than 30 minutes time out. Add filters, split the query into smaller chunks, or test with LIMIT.
  • Result size: result sets over 100 GB may fail or take a long time. Select only the columns and rows you need, or page with LIMIT and OFFSET.
  • Memory: large joins, window functions over huge datasets, and complex aggregations may run out of memory. Filter early, use smaller time windows, or pre-aggregate with a Query.

Best practices

  • Test your query in the Explorer first.
  • Start with LIMIT to check the query before processing large datasets.
  • Cast timestamp and decimal columns explicitly when using Parquet.
  • Use Parquet for analytical workloads and CSV for maximum compatibility.
If you run into an issue not listed here, contact support with your SQL query, the error message, your cloud (Nekt Express, GCP, or AWS), and the output format.