Upload a File
In this article
In this article group
If you have your data in a CSV, JSON or JSONL file, you can import it into a collection from the dashboard, without writing any code. Files can be up to 20 MB.
Into a new collection
Open your Cluster Dashboard and click on "New Collection".
Under "Upload a file", click on "Choose a file" (or drop the file there).
We'll suggest a schema based on the first 1,000 rows. Review the field types and facets.
Click on "Create collection and import N rows".
Into an existing collection
Open the collection, and click on "Add Documents".
Under "Option 1: Upload a file", choose your file.
Pick the "Action Mode":
create,upsertorupdate.Click on "Import N rows".
For a CSV, the first row needs to be a header row with the field names. If a column is named id, its values are used as the document IDs.
If a batch fails, click on "Retry from batch N".
If your documents don't have an id field, retrying a batch can create duplicate documents. So you want to include an id in your data if you think you'll need to retry.
For files over 20 MB, you want to use a client library or curl, or ask your AI agent through the MCP server.