All nodes/Data/DataFrame I/O

Write CSV

Writes a DataFrame to a CSV file in the connected file_store. Overwrites the target file if it exists.

Try it

When to use it

Saves a table as a CSV file in the store. The neighbouring File Write writes text and can replace individual lines; here the text is assembled for you out of the table — delimiter, encoding, header row — and the file is always replaced whole.

Pick CSV when a machine will read the file or move it into another system. If a person will open it in Excel, look at Write XLSX instead: no encoding or delimiter questions there.

How it works

The “DataFrame” port carries both the trigger and the data — the node has no separate execution input. If another branch should decide the moment of writing, enable “Split input ports” and you get separate legs for execution and value.

The input takes more than a table: a list of objects (JSON records) and CSV text are parsed by the node itself, so a converter before the write is not always needed.

The path is a template: reports/{{ variables.session_id }}.csv gives one file per session. The output is path, bytes_written and rows_written (data rows, header excluded).

Common mistakes

• Expecting an append. There is no append mode: every run replaces the file. History is made with a path carrying a date or a session id.
• The WebDAV folder does not exist. The workspace store creates missing folders itself, WebDAV does not: writing reports/2026/x.csv fails until the directory exists.
• Cyrillic or emoji in cp1251. A character the encoding cannot carry fails the step at encode time, and retrying does not help. For a file people will double-click in Excel, choose “UTF-8 with BOM”.
• An empty table. An empty result is not an error: you get a file with the header alone and rows_written: 0. If an empty report is meaningless, check for it earlier in the graph.

Inputs

Outputs

Configuration