AI Nexus releases Structured Data Editor for VS Code

The extension comes out of a specific friction the author describes: text files are increasingly the raw material for LLM work — Markdown tables, JSON, and CSV used in RAG and prompt test scenarios — but editing those structured formats by hand in a code editor is painful. The common workaround has been to copy data into Excel or a spreadsheet, which the author says repeatedly caused trouble: codes like 01234 being turned into 1234, dates reformatted, character encodings and line breaks broken on save, and quote or comma parsing errors that damaged AI evaluation data.

The design choice that stands out is a refusal to auto-repair. When a CSV has ragged rows — for example, a Japanese prompt-evaluation file where an unquoted thousands separator split a column — the tool does not try to infer which commas belong to sentences and which are delimiters. Instead it halts, reports the line number, and lets the user jump there. That is a deliberate trade of convenience for safety, on the view that a wrong guess would silently corrupt the user's data.

Because the source stays plain text, changes remain trackable in Git, and the extension acts as a visual layer only when editing is needed. For teams already keeping prompts and evaluation sets in version-controlled text files, that combination — spreadsheet-like editing without leaving the file format — is the practical draw.