Jev is a model from TypeSafe designed for structured decisions: choosing an option, evaluating a yes/no question, or scoring against a rubric. In Magic CSV, those decisions become cells beside your original data. You define what an answer can be; Jev evaluates each row using the columns you select.
What the result looks like
Illustrative example with synthetic data and authored answers, not a recorded model run. Your results will depend on the source data and instructions.
Scroll tables sideways to see every column.
| # | Customer message |
|---|---|
| 1 | Can you send a corrected invoice? |
| 2 | Thanks, that fixed it! |
| 3 | I still cannot log in. Please help. |
| # | Customer message | Needs a reply? |
|---|---|---|
| 1 | Can you send a corrected invoice? | Yes |
| 2 | Thanks, that fixed it! | No |
| 3 | I still cannot log in. Please help. | Yes |
Try it, step by step
Open your file
Open the Magic CSV workspace and import a CSV or Excel file. Start with a small, representative file so you can check the answers before processing a larger export. The downloadable sample has three customer messages.
Ask one question
Choose AI actions → Flag. Name the column “Needs a reply?” and keep Yes / No as the answer type. Under Read these columns, select Customer message. In More guidance → Extra instructions, enter: “Yes when the customer asks a question or describes an unresolved problem. No for thanks or confirmations with no remaining request.”
Fill and check the column
Run the column and watch answers arrive in the sheet. Click an AI cell to inspect or correct its answer. Review flagged cells and check a few unflagged ones too. A model can be confident and still be wrong.
Build on the answer
Add another AI column for a separate decision, such as Topic or Urgency. Later columns can read earlier AI answers. Keep your original message column, then export to CSV or Excel when the results are ready.
Three ways to use Jev in a spreadsheet
Categorize chooses from your allowed answers. Use it for expense categories, product types, feedback themes, or request routing. Add an Other option when a row might not fit. A fixed list keeps labels consistent across the file; it does not guarantee the selected label is correct.
Flag answers a focused yes/no question: “Needs a reply?”, “Mentions a cancellation?”, or “Missing delivery details?” Score places a row on a 0–100 scale using the descriptions you enter for 0 and 100. Use separate columns for separate judgments so the results remain understandable.
Why this is different from asking a chatbot about a file
The output is attached to the row it describes. You can review an individual answer, define a column’s instructions, or use one decision as input to the next. You do not have to copy labels out of a conversation and match them back to the source rows.
Jev handles the constrained decision columns. Magic CSV also includes AI text transformations for extracting fields, rewriting text, composing columns, and expanding rows; those generative operations use OpenAI. Local cleanup, editing, and file combining live in the same workspace. Choose the tool that matches the job.
What to expect before you run it
Magic CSV reads the fields you provide. It does not browse company websites, look up missing facts, or verify statements in a customer message. Supply the context needed for the decision, exclude irrelevant personal fields, and give ambiguous inputs a sensible fallback.
Answers stream as the service returns them. File size, question complexity, provider latency, and request limits affect the time it takes; a whole file is not guaranteed to finish instantly. Current AI sheets support up to 1,000 rows, 20 AI columns, and 100 columns in total.
Files are parsed in your browser. For AI actions, selected source fields and instructions are sent through the server to the relevant model provider. Magic CSV is an independent application using TypeSafe Jev, rather than an official TypeSafe spreadsheet product.
Common questions
- Do I need a Jev API key?
- No. Magic CSV supplies the integration. Its subscription includes the AI tools, so you do not need to write code or bring a TypeSafe API key.
- Is this an Excel or Google Sheets add-in?
- No. Magic CSV is a browser workspace. Import a CSV or Excel file, work on it, and export the result. It does not automatically sync changes back to an open workbook or Google Sheet.
- Can Jev write text or formulas in my spreadsheet?
- Jev decision columns produce categories, yes/no answers, or scores. For new text, use Magic CSV’s Compose a column or Rewrite or format actions, powered by OpenAI. Magic CSV does not calculate spreadsheet formulas.