Classify text in Google Sheets with Jev

How do you classify text in Google Sheets with Jev?

Use one question to identify the feedback topic and another to suggest when a person should review it. The two answers describe different things: a resolved bug can remain a Product defect while receiving Review later.

Our six messages are original synthetic examples. Try them below, or use the Playground for another small classification task. The web demo makes real API requests; it does not execute Google Sheets formulas.

The formula interface comes from dbredesen's Jev for Sheets. We inspected its source at commit 6644d9544f8e8f15909f80555b18a3a8b1cd4425. We did not independently install its Apps Script project or reproduce the author's worksheet tests.

Original schematic showing two Jev formulas and a separate request containing two Choice questions — Original schematic: the source formulas and our web demo use different request paths.

Keep topic and next step as separate questions

The complete rules JSON defines both questions and their shared instructions. Use stated events and timing. A customer's report is a claim to investigate, and tone alone does not establish urgency.

The topic question includes Billing, Product defect, Feature request and Other. It uses Other when there is no clear main topic. The next-step question has Review now and Review later; neither label authorizes an automatic refund, reply or rejection.

Review now covers a stated current block on an essential task, an unresolved incorrect or duplicate charge, or ongoing data loss.
Review later covers routine feedback, ideas, resolved issues and unclear current impact. Unclear cases stay in the normal queue for clarification.

Original schematic of four topic labels and two next-step labels for customer feedback — Original schematic: topic and next step each have their own question and allowed answers.

Set up the author's formula project

The author's installation instructions describe a script attached to one spreadsheet. Execution happens on Google's servers. These are documented steps, not our installation test. The reviewed version does not have an approved public Marketplace listing.

Clone the repository, check out the reviewed commit, and use Node 20 or newer. Run npm test, then npm run build.
Open a new blank sheet and choose Extensions > Apps Script. Replace Code.gs with dist/Code.gs from the build.
In Project Settings, expose the manifest and replace it with dist/appsscript.json. Save and reload the sheet; this path requires no deployment.
Run all four registration checks in the README before connecting a key. An Unknown function error means registration failed.
Open the project's API key & connection menu, authorize it, save your TypeSafe key, test it and connect it to this spreadsheet. Keep the key out of cells.

Add two JEV_CHOICE formulas

Import the sample feedback CSV and put its text column in A2:A7. Create a Rules tab and paste the one-row Rules TSV into A1; it fills A1, B1 and C1 with the full shared instruction and two questions. The ready-to-copy formulas use those cells. Both downloads use the same rules as our demo.

For the topic column, enter =JEV_CHOICE(A2,Rules!$A$1&CHAR(10)&Rules!$B$1,"Billing","Product defect","Feature request","Other").

For the next-step column, enter =JEV_CHOICE(A2,Rules!$A$1&CHAR(10)&Rules!$C$1,"Review now","Review later"). Fill both formulas down through row 7.

These examples follow the reviewed function signature. Each question includes the full policy. The source adds instructions scoped to data and sends the cell value there. We have not executed these formulas in Sheets, so their results are not assumed to match our web run.

Try the original two-question web demo

Load the six samples and keep both dimensions selected. They cover duplicate charges, a blocked export, a feature idea, praise, a resolved glitch and mixed feedback. Run them, read the labels beside the unchanged text, and export the results.

Then edit one message and rerun it. For example, change whether the reported problem still blocks work. This creates a new observation under the same rules; it does not establish how the model handles every customer message.

Classify the sample feedbackLive API

Turn a table of text into useful decisions.

Classify each text row using your instructions and choices. Review every result, then export the labels and summary.

Inspired by dbredesen on Jev for Sheets. Inspired by Jev for Sheets and its typed formula workflow. These six customer-feedback rows and the two Choice rules are original to this site. This web demo sends two independent questions together per row; it does not run Google Sheets formulas, reproduce the author's worksheet tests, or establish the author's performance.

Choose a text fileCSV, TSV or TXT · up to 128 KB

Original sample · 6 rows

Input format

Or paste your text

Text column6 rows · maximum 20

Row	Text to classify
1	Two identical subscription charges appeared on my card today. I bought only one plan, and both charges are still posted. Please investigate the duplicate payment.
2	The CSV export button returns an error every time I press it this morning. Exporting is the only way I can send today's required report, so that task is blocked right now.
3	Could you add a weekly email digest with a chart of completed tasks? The dashboard works as documented; I would simply like a scheduled summary.
4	The new navigation is easy to understand, and I found the setup guide quickly. Thanks for making the first visit straightforward.
5	Yesterday the date filter briefly showed the wrong month. Refreshing fixed it, and I have completed my work. I am reporting the resolved glitch for your records.
6	The receipt looks higher than I remember, and yesterday's export looked incomplete. I am not sure whether either problem is still happening or which one matters more. Please check both when you can.

Up to 1,000 characters per row and 12,000 in total. Empty rows are skipped. Only the selected text column is sent when you run; files are not stored by this site.

Classification instructions

Choose what to classify

1–3 dimensions · 2–8 choices each

Topic

Billing · Product defect · Feature request · Other

Customize dimension

Dimension labelQuestionChoices, separated by commas

Next step

Review now · Review later

Customize dimension

Dimension labelQuestionChoices, separated by commas

Jev selects from your choices. Totals and percentages are calculated by the app, rather than generated as a summary.

One run · up to 60 classification decisions

Edit any input to explore a different decision. No request is sent until you run it.

Your next decision

Not run yet

Load the sample or choose a file, then run your classification. Results and totals will appear here.

Results are not stored by this site. Your submitted text is processed by the API provider.

How does the web request differ from the formulas?

Our adapter puts one row in state.text. It builds two Choice questions with the TypeSafe JavaScript SDK. Each question's instructions contain its task, the shared classification rules and a scope statement limiting the judgment to that text.

Both questions share one systemOne call. In the author's wrapper, two formulas make separate calls using state.data. Matching business rules does not make these requests identical. Our Jev model guide explains the broader decision interface.

What did the six-row run return?

On September 25, 2026, all six requests in our controlled adapter run succeeded, returning 12 choices from jev-1.13.0. Download the observed results to inspect each input and response.

Median client-observed duration was 348.5 ms per request, ranging from 296 to 902 ms. These adapter timings include network and response handling. They exclude browser interaction and Turnstile, and say nothing about Apps Script execution.

The usage-derived estimate was $0.000233016 for those six requests using the recorded model rate. It excludes separate selection and review calls and is not an invoice. No expected labels were independently annotated, so we report outputs rather than accuracy.

Recorded labels for six synthetic rows

Recorded labels for six synthetic rows
Sample	Topic	Next step
feedback-01	Billing (100%)	Review now (100%)
feedback-02	Product defect (100%)	Review now (100%)
feedback-03	Feature request (100%)	Review later (100%)
feedback-04	Other (100%)	Review later (100%)
feedback-05	Product defect (100%)	Review later (100%)
feedback-06	Billing (57%)	Review later (95%)

Percentages are selected-label probabilities, not measured accuracy. All expected labels were null; this is an unscored synthetic smoke test.
Measured through our server adapter, not through Google Sheets. The source project installation was not independently tested.

Original six-row API run, jev-1.13.0, recorded 2026-09-25.Verified 2026-09-25

Inspect the mixed feedback before acting

The sixth message mentions a possibly high receipt and an incomplete export, without identifying a main concern. Jev selected Billing at 0.57; Other received 0.22 and Product defect 0.21. Topic confidence was 0.42. Its next-step result was Review later at 0.95.

A separate same-model review, explicitly matching each row to each question, marked 11 decisions consistent and this topic decision ambiguous. Ambiguous received 0.60 in that review. We retained the original Billing output; the review is another model judgment, not independent validation.

The Choice reference distinguishes the selected label, its probability distribution and confidence. Neither probability nor confidence measures accuracy on your customers. Read the feedback before turning a label into work.

Original schematic connecting the mixed feedback result to human review — Original schematic: preserve the returned label and review the ambiguous message in context.

What should you check before using real feedback?

The reviewed source adapter has no cache, cross-formula batching or automatic retry. Two formulas evaluated once on six nonblank rows imply 12 requests. Recalculation can add requests; fewer calls alone do not prove lower token use or cost.

Google limits custom functions to 30 seconds. The source stores a connected key for the spreadsheet; collaborators consume that connection's usage. Trust editors who can modify the attached script, and remove private customer details before testing.

Google also offers AI() categorization. This tutorial explores one Jev workflow without benchmarking alternatives. Keep a copy of your text and rules, inspect a small sample, then adapt the pattern using the examples library.

Questions about spreadsheet classification

Does JEV_CHOICE return confidence?

It returns the selected label. The author's raw JEV function returns full response JSON, including probability fields. Our web demo also retains the distributions.

Why do two formulas make more requests than the web demo?

Each source formula makes its own request. Our demo sends both questions together for one row. Six rows still require six requests; uploading them does not create one combined call.

Can these formulas classify an entire range at once?

The reviewed convenience functions accept one data cell or literal. Vertical ranges can supply allowed labels, but a multi-cell data range is rejected. Fill the formulas down instead.