JSON Lines (JSONL) Viewer & Converter

JSON Lines (also called JSONL, and nearly identical to NDJSON) stores one complete JSON value per line, separated by a newline. There is no enclosing array and no commas between records. That small change makes a big difference in practice:

• Streaming. A program can read, process and discard one line at a time, so a 20 GB file never has to fit in memory.
• Appending. Writers add a record by appending a line, without rewriting a closing bracket. That is why structured logs from tools such as Docker, Bunyan, Pino and many cloud logging exports use it.
• Parallel work. Files can be split on newlines and processed on many machines, as BigQuery, Spark and Hadoop do for newline-delimited JSON imports.
• Machine learning datasets. Fine-tuning files for OpenAI, Anthropic Bedrock and Hugging Face datasets are JSONL, with each line holding one training example such as a messages array.

Each line is parsed on its own with a strict JSON parser. A line that fails does not stop the others: it is listed with its line number and the parser's message, so you can jump straight to it in your editor. Blank lines are skipped, a trailing newline at the end of the file is expected, Windows CRLF line endings are accepted, and a UTF-8 byte order mark at the start is removed.

The Table view uses the union of top-level keys from all records as columns, in the order they first appear, and shows nested objects and arrays as compact JSON. It displays the first 200 rows so very large files stay responsive; the exports always include every record. The Record view pretty-prints one record at a time, which is easier for deeply nested chat transcripts.

Conversions: JSON array wraps all valid records in brackets. CSV follows RFC 4180: fields containing commas, quotes or newlines are quoted and inner quotes doubled, rows end with CRLF, and records that are not objects go in a value column. To go the other way, paste a JSON array and choose Convert JSON array to JSONL.