Upload files

You can upload files directly into Querio without connecting an external database.

Uploaded files are converted to Parquet and stored in the Files/files/ folder. Querio also represents them in a managed datasource called uploads, where you can manage metadata and agent access.

Supported formats

• CSV
• TSV
• Parquet
• JSON
• Excel (.xlsx)

For Excel workbooks, Querio imports the first worksheet and uses its first row as column headers. Other worksheets are not imported. To upload another worksheet, save it as a separate workbook. Legacy .xls files are not supported; save them as .xlsx first.

How to upload

1. Open the Context page for your workspace
2. Click the upload button in the Datasources header of the sidebar
3. Choose one or more files using the file input
4. Click Upload. Querio validates and ingests each file on the server and displays any errors. Each successful upload creates a Parquet file named after the original file.

Each file appears in the uploads datasource as a catalog table so you and the agent can see its structure and metadata. It is not stored as a table in DuckDB.

Upload from an automation

You can also send files to the uploads datasource from a scheduled job or other automation. Create a programmatic access token in Querio with the Import data scope, then send the file as multipart form data:

curl -X POST \
  "https://app.querio.ai/api/v2/workspaces/<workspace-id>/datasources/uploads/tables" \
  -H "Authorization: Bearer <your_programmatic_access_token>" \
  -F "files=@/path/to/your-file.csv"

The token must belong to the workspace in the URL. You can include more than one files field in a request; Querio processes them using the same limits and ingestion flow as uploads made in the app.

Files that cannot be parsed are rejected rather than imported with invalid rows skipped.

Limits

• 100MB per file
• 100MB total per batch
• 10 million rows per file
• 1,024 columns per file

After uploading

Once ingestion is complete, the file appears in the Files/files/ folder and is represented in your Catalog under the uploads datasource. You can query it in an Explore session with the local DuckDB engine using the Parquet file path:

SELECT * FROM "files/your_csv.parquet"

If Querio detects that the data structure or column descriptions have drifted from what the catalog says, it will flag this and suggest an update.

Notes

• Only one file can be written to the uploads datasource at a time per workspace. If multiple files are uploaded at once, they are ingested sequentially.
• File uploads are available for workspace v2 (fs-mode) workspaces.