Your data is read and written on this device. Nothing is uploaded. Turn a spreadsheet of questions and answers into a JSONL training file, or check that an existing JSONL file is valid before you use it. It is free and private.
How to build a dataset
- Open the Build a dataset tab and choose the format: Chat or Completion.
- Press Open CSV or paste into CSV text. Try an example loads sample rows.
- Tick First row is a header, then choose the Prompt column, Answer column and, if you have one, System column.
- Add more rows with Add example if you want.
- Check the counts for examples, estimated tokens, errors and warnings.
- Press Copy or Download .jsonl.
How to check a JSONL file
- Open the Check a JSONL file tab.
- Press Open .jsonl or paste into JSONL text.
- Set the Expected format, or choose Work it out from the file.
- Read how many records there are and how many are without errors.
- Press Download report CSV for the full list of problems.
Chat and completion formats
Chat datasets have one JSON object per line with a messages list: an optional system message, then user and assistant turns, ending with the answer you want the model to learn. Completion datasets have a prompt and a completion on each line. Most current fine-tuning services expect the chat format; some evaluation tools and older services use prompt and completion.
What the checker looks for
Lines that aren't valid JSON, lines that aren't one object, missing or empty messages, unknown roles, empty content, examples with no assistant answer, system messages in the middle, the same role twice in a row, tool messages without a tool_call_id, mixed formats, duplicate lines and blank lines. Errors usually make an upload fail; warnings are worth a look but may be fine.
Tips for a good dataset
- Make the answers exactly the style you want the model to copy.
- Keep the examples consistent. Mixed styles teach mixed behaviour.
- Remove duplicates and mistakes before you build the file.
- Keep a separate set of examples back to test with afterwards.
- Never include private data you would not want to appear in an answer.
Who uses it
Developers and small teams preparing training examples, analysts converting a spreadsheet into JSONL, and freelancers delivering data for a client. To estimate cost, try the AI cost estimator, and to test answers use the AI answer schema check.
Questions
How should my CSV look?
One example per row, with a column for the prompt and one for the answer, and optionally a column for a system message. Columns named prompt, question or input and completion, answer or output are picked up automatically; you can change the choice from the lists.
How many examples do I need?
Providers usually ask for at least ten, and results tend to improve with fifty to a few hundred clear, consistent examples. Quality matters more than quantity.
Are the token counts exact?
No, they're estimates at about four characters per token plus a few tokens for each message, which is close enough to judge cost and spot examples that are far too long.
Can it check a very large file?
Yes. Files with tens of thousands of lines check in a moment. The report lists the first 1,000 problems on screen and every problem in the CSV.