All nodes/Data/DataFrame I/O

Convert

Universal payload converter: json ↔ dataframe ↔ csv. Choose the source and target type.

Try it

When to use it

Moves a value between three forms: JSON (an object or a list of objects), CSV text and a DataFrame table. Most often the node is needed on the way out of the table world: a table cannot go into a prompt, an HTTP body or a Template — it has to become JSON or CSV first.

The other direction needs it less. The DF family and both writers (CSV, XLSX) already coerce a list of objects or CSV text into a table themselves. Add the converter explicitly when the parsing has to be spelled out — a ; delimiter, say, or a different header row.

How it works

There are four conversion pairs: JSON → DataFrame, DataFrame → JSON, DataFrame → CSV, CSV → DataFrame. There is no direct route between JSON and CSV: that takes two nodes in a row, through DataFrame. Graph validation says so before the run.

The fields follow the chosen pair: “CSV delimiter” appears when CSV is on either side, “CSV include header” when producing CSV, “CSV header row” when reading it, and “Max rows” only when producing JSON.

When “From” and “To” match, the value passes through unchanged.

Common mistakes

• “From” does not match what actually arrives. The field is a declaration, not detection — the node does not sniff the format. A string with “From: JSON” produces an error along the lines of “Cannot convert str to dataframe”.
• JSON → DataFrame from a list of non-objects. A list of objects (or a single object) is expected; a list of numbers or strings will not become a table.
• CSV → DataFrame with the wrong delimiter. The default is ,; a ; export collapses into a single column, with no error raised.
• DataFrame → JSON without “Max rows” in front of a model. The whole table goes into the prompt and is paid for in tokens. Cap the rows, or convert to CSV — it is noticeably more compact.

Inputs

Outputs

Configuration