JSONL Formatter — Free Online JSON Lines Validator & Viewer

JSONL Formatter is a free, browser-based tool that validates, formats, and inspects JSON Lines (JSONL / NDJSON) files line-by-line. Paste your data, instantly see which lines are valid or broken, expand any line for a pretty-printed view, and download only the clean output — all without sending a single byte to a server.

The Problem

JSON Lines (.jsonl) is everywhere in modern data pipelines: OpenAI fine-tuning datasets, LLM training corpora, streaming API responses, log files, ClickHouse exports, and NDJSON feeds. The format is deceptively simple — one JSON object per line — but when things go wrong, they go wrong silently.

A single malformed line in a 10,000-line fine-tuning dataset will fail the upload with a cryptic error. Log aggregation pipelines drop records silently when one line is invalid. Debugging means cat file.jsonl | python -c "import sys,json;[json.loads(l) for l in sys.stdin]" — fragile, slow, and opaque.

There was no dedicated, clean browser tool for this. Every workaround was either a CLI command you have to Google every time, a full-featured IDE plugin you don't want to install for a one-off check, or an upload-based service you can't trust with training data.

How It Works

1. Paste and Validate Instantly

Paste any JSONL or NDJSON content into the input area and click Validate & Format. Each line is parsed as an independent JSON object — exactly how a real JSONL parser would process it. Results appear immediately.

2. Line-by-Line Status

Every line gets a status badge:

• Green check — valid JSON object
• Red X — invalid, with the exact parse error message (Unexpected token 'n', Expected ':' but found ',', etc.)
• Grey dash — empty line (skipped without noise)

A stats panel at the top shows the total line count alongside the valid, invalid, and empty counts at a glance.

3. Expand to Pretty-Print

Click any valid line to expand a fully-indented, readable view of the JSON object. Useful when you need to inspect a specific record without copy-pasting into a separate JSON formatter. Click again to collapse.

4. Export Clean Output

Two export options:

• Copy valid — copies only the valid lines to clipboard as compact JSONL, ready to paste or re-upload
• Download .jsonl — saves a valid-lines.jsonl file to your machine

5. Sample Data

The Load Sample button inserts a short demo JSONL snippet — including one intentionally broken line — so you can explore the tool without needing your own data.

Key Features

• Line-by-line validation — each line parsed independently, not as a whole document
• Exact error messages — know precisely what is wrong and on which line
• Expand / collapse per line — pretty-print any record inline
• Stats panel — total / valid / invalid / empty counts
• Copy valid lines — clipboard export of clean output
• Download .jsonl — save valid lines as a file
• Expand all / Collapse all — bulk toggle for large datasets
• Sample data — one-click demo with a broken line included
• 100% client-side — no uploads, no server, no data retention
• Dark mode — respects system preference, switchable in header

Technical Implementation

Core Technologies

• Next.js 16 with App Router
• TypeScript in strict mode
• Tailwind CSS v4 with OKLCH color tokens
• shadcn/ui components (new-york style)
• Framer Motion for animated expand/collapse transitions
• Lucide React icons

Architecture

The entire JSONL parser is a single pure function:

function parseLines(input: string): ParsedLine[] {
  return input.split("\n").map((raw, index) => {
    const trimmed = raw.trim();
    if (trimmed === "") return { index, raw, status: "empty" };
    try {
      const parsed = JSON.parse(trimmed);
      return { index, raw: trimmed, status: "valid", parsed };
    } catch (e) {
      return { index, raw: trimmed, status: "invalid", error: e.message };
    }
  });
}

Each line is split and individually fed to JSON.parse. The native JSON parser's error message is surfaced directly — no custom error formatting that might hide information.

Line expand/collapse state is managed with a Set<number> so toggling one line doesn't re-render others. The list uses AnimatePresence from Framer Motion for smooth height transitions when expanding.

Export uses the browser's native Blob + URL.createObjectURL pattern — no dependencies, no file size limits beyond available RAM.

Use Cases

AI / ML Training Data

OpenAI, Anthropic, and Mistral fine-tuning APIs all accept JSONL. A single malformed line breaks the entire upload. Paste your fine-tuning file here, confirm all lines are valid, then download the clean output to upload with confidence.

LLM Prompt Dataset Curation

Prompt datasets (ShareGPT format, Alpaca format, instruction-tuning corpora) are distributed as JSONL. Validate and inspect records before feeding them into a training pipeline.

Log File Debugging

Structured logging libraries (winston, pino, structlog) write JSONL to stdout. When a deployment goes wrong, paste the log output here to quickly identify malformed entries.

NDJSON API Responses

Some streaming APIs (Elasticsearch _search?scroll, Kibana exports, ClickHouse HTTP interface) return NDJSON. Paste the response and browse individual records without writing a script.

Data Pipeline Validation

Before loading a JSONL file into a database or message queue, validate it here to catch type mismatches, missing fields, or truncated lines that would fail silently downstream.

Why JSONL Formatter?

vs. python -m json.tool

python -m json.tool only handles standard JSON, not JSONL. You'd need a loop: while IFS= read -r line; do echo "$line" | python -m json.tool; done — slow, verbose, and it stops at the first error.

vs. jq

jq is powerful but requires installation and CLI familiarity. cat file.jsonl | jq '.' also stops at the first invalid line. There's no line-by-line summary of what passed and what failed.

vs. Online JSON Formatters

Standard JSON formatters reject JSONL because it isn't valid JSON as a whole. They try to parse all lines together and throw a top-level syntax error.

vs. IDE Plugins

VS Code's JSON tools don't understand JSONL natively. You'd need a specific extension, and it's overkill for a one-off validation task on data you'd rather not save to disk.

Try it now: jsonl-formatter.tools.jagodana.com