Debugging JSONL & NDJSON Streams Without Uploading Logs
Fine-tuning datasets, server structured logs, and batch AI evaluation results use .jsonl / .ndjson because append-only line streams survive mid-write crashes far better than giant [...] JSON files. Unfortunately, a single truncated line causes standard JSON formatters to throw a syntax error on line 8,421 without showing the valid events around it.
This zero-upload viewer parses line-by-line right in your browser tab, skips bad lines while reporting exact line numbers and error reasons, lets you search instantaneously across formatted JSON cards or a flat columnar table, and exports filtered matches to JSON or CSV.
Workflow & Workstation jq Equivalents
- Drop a
.jsonl,.ndjson, or.jsonfile above, or click Try sample .jsonl (40 events). - Type a model name, status, or error key into Quick filter to narrow both formatted JSON cards and the tabular preview simultaneously.
# Pretty-print failed records from a huge local JSONL file with jq
jq -c 'select(.ok == false)' events.jsonl | head -n 20
# Convert JSONL to a clean JSON array in shell
jq -s '.' events.jsonl > events.jsonBoundaries & Schema Evolution
- Sparse columns: When line #12 introduces a new JSON key absent from earlier lines, the flat table discovers the union of all keys across the loaded records so columns never drop silently.
- Browser RAM protection: Loading stops gracefully at 10,000 records so multi-hundred-megabyte log archives stay responsive.
Frequently Asked Questions
What is the difference between JSONL / NDJSON and a standard JSON array?
JSONL (Newline-Delimited JSON) stores one standalone JSON object per line rather than wrapping records in [...] brackets. It allows append-only streaming in LLM datasets, OpenTelemetry logs, and BigQuery exports.
What happens if some lines in my log file are malformed or half-written?
Instead of failing on the whole file like JSON.parse, this viewer isolates invalid lines, reports exact 1-based line numbers and parser errors at the top, and continues rendering every valid record.
Can I flatten nested JSONL keys into a downloadable CSV?
Yes. Switch to Table Preview and click Export CSV—top-level keys become columns while nested sub-objects serialize cleanly as JSON strings inside quoted cells.