Wapgee Logowapgee

Tools / JSON / JSONL

JSON ⇌ JSONL

Paste a JSON array or JSON Lines on the left. The tool detects the format, converts both ways, and lets you copy or download the result. Built for LLM fine-tuning files and batch APIs that expect one JSON object per line.

InputJSON
JSONL

JSON arrays and JSON Lines

A JSON array is one document: opening bracket, comma-separated values, closing bracket. JSON Lines (JSONL, also called NDJSON) is the same values written one per line with no wrapping array. That shape streams well, appends without rewriting the file, and is what OpenAI and Claude batch APIs expect for request files and what many fine-tuning pipelines use for training examples.

JSON
[
  { "prompt": "hi", "completion": "hello" },
  { "prompt": "bye", "completion": "goodbye" }
]
JSONL
{"prompt":"hi","completion":"hello"}
{"prompt":"bye","completion":"goodbye"}

Auto-detect and both directions

Paste either form on the left. If the whole input parses as JSON, the tool converts to JSONL (one compact value per line). If it does not, but individual lines parse, it builds a JSON array. Use Direction to force JSON → JSONL or JSONL → JSON when auto-detect is wrong for your file. Pretty JSON controls indentation on the array output; JSONL is always compact.

Blank lines are ignored. CRLF and LF both work. Trailing newlines are fine. Invalid JSONL lines show their line number; turn on Skip bad lines to keep the rest of the file.

Fine-tuning and batch files

When a provider asks for a .jsonl upload, start from the JSON array you already have in tests or fixtures, convert here, then download. Going the other way is useful for inspecting a batch result file as a single pretty-printed array in an editor. Everything stays in the browser, so corpora with private prompts are safe to paste.

FAQ

Is my data uploaded?

No. Conversion runs in your browser. Nothing is sent to a server, so batch files and fine-tuning corpora stay on your device.

What is JSONL used for?

JSON Lines (also called NDJSON) stores one JSON value per line. OpenAI and Claude batch APIs, LLM fine-tuning datasets, and many log pipelines use this format because it streams and appends cleanly.

What happens on invalid lines?

JSONL to JSON reports the line number of each bad line. Turn on Skip bad lines to keep the valid rows and drop the rest.

  • Token Counter & LLM CostPaste a prompt, see token counts and estimated cost across Claude, GPT, and Gemini models side by side, before you send it.
  • RAG Chunk VisualizerPaste a document, set chunk size, overlap and strategy, and see exactly where it splits, with token counts, problem flags and JSON or JSONL export. Nothing leaves your browser.
  • LLM Tool Schema BuilderDefine a tool in a form and copy correct JSON for Claude and OpenAI function calling, plus raw JSON Schema, all in the browser.
  • JSON Schema → TypeScript / ZodPaste a JSON Schema and get TypeScript interfaces plus Zod validators: optional fields, enums, $refs, and more, all in-browser.
  • HTML Table → Markdown / CSVPaste HTML with one or more tables and get clean GitHub-flavoured Markdown or CSV, with colspan, rowspan, header rows and links handled in your browser.
  • CSV ⇌ JSON ConverterPaste or drop CSV and get JSON, or the other way around, with quoting, delimiters, headers, and type inference handled in the browser.
  • JSON to SQLPaste a JSON array of objects and get CREATE TABLE plus INSERT for PostgreSQL, MySQL, or SQLite, with inferred types you can override in the browser.
  • CSV to ParquetTurn a CSV into a Parquet file in the browser. Types are inferred, nulls stay optional, and you can override each column before download.
  • Parquet to CSVOpen a Parquet file to see its schema, row groups, and a preview, then export CSV or JSON. Large files are read in ranges, and nothing is uploaded.
  • Parquet ViewerOpen a Parquet file and view the first 100, 150, or any number of rows with the schema and row count. Read in your browser, nothing uploaded.
  • HTML to MarkdownTurn any page or pasted HTML into clean Markdown, perfect for docs, README files, and preparing web content as LLM context.
  • Markdown to HTMLPaste Markdown and get clean, formatted HTML with a live preview. Tables, task lists, and fenced code render just like GitHub.
  • Markdown to PDFPaste Markdown and download a PDF of the rendered page. Tables, task lists, fenced code, Mermaid diagrams and LaTeX math, all in the browser.
  • JSON ⇌ YAML ⇌ TypeScriptConvert config and API payloads between JSON, YAML, and TypeScript interfaces in one place, with auto-detected input.
  • URL Parser & BuilderInspect every URL component, edit query params in a table, strip tracking, and copy a correctly re-encoded link, all in-browser.
  • Markdown Table GeneratorEdit cells in a spreadsheet-style grid and get a perfectly aligned Markdown table, with alignment, headers, and CSV import.
  • Diff CheckerCompare two blocks of text or code and see exactly what changed, line by line, side by side or unified, whitespace-aware.

The weekly download

One new tool a week. No spam.