SCENARIO / REVIEWS
Classify review topics and sentiment separately with Jev. Compare mixed, positive and unclear examples, edit a JSON template or analyze your own reviews.
One review can mention several issues while expressing positive, negative, neutral or mixed sentiment. Topic labels and sentiment are independent decisions; unknown sentiment stays open for review. The tool uses fixed themes rather than automatic topic discovery.
Positive, negative and missing-evidence examples follow. Authored answers do not change when you edit a request. Read the practical guide
Topics describe reported issues; sentiment describes the expressed evaluation. Mixed preserves both praise and criticism. Neutral means factual without evaluation; unknown means no supported judgment. These are authored teaching examples, not live model outputs.
| Review | Topic labels | Sentiment | Why |
|---|---|---|---|
| Keeps drinks warm, but the lid leaks. | Product quality | Mixed | Praises warmth and reports an unwelcome leak. |
| Love the color! | Insufficient information | Positive | Expresses praise without concrete issue evidence. |
| The lid leaks. Very disappointed. | Product quality | Negative | Explicitly expresses dissatisfaction. |
| The cup holds 500 ml. | Insufficient information | Neutral | States a fact without an evaluation. |
| How much does this cup cost? | Insufficient information | Unknown | Asks a question rather than giving an evaluation. |
The workspace asks separate topic and sentiment questions in one Jev request, still one credit per item. Filter sentiment, correct results and export; original model suggestions remain alongside your edits. Older projects without sentiment show “Not assessed” and do not automatically rerun.
Open the sentiment analysis tool →Who this is for Product and commerce teams organizing feedback with original evidence.
Authored examples · No live inference{
"review": "It keeps drinks warm, but the lid started leaking after two days. The delivery box was crushed; the cup itself was not broken."
}The leak is a quality issue and the crushed box is a packaging issue. The cup was not reported broken. Sentiment is mixed: warmth is praised while leakage and packaging are criticized.
Next action: Count both labels and retain the original for review.
{
"review": "One month of use, no leaks, easy to clean, packaging intact."
}The text denies the issue. Mentioning the word leak is not evidence of a leak. Sentiment is positive: easy cleaning and satisfactory use express approval.
Next action: Keep as no reported issue; exclude from issue counts.
{
"review": "How much does this cup cost?"
}No specific defect, delivery or usability problem is described. Sentiment is unknown: the text does not offer a clear product evaluation.
Next action: Ask a person to inspect context; do not force a category.
Edit the material and question criteria to create a request. Stable field names stay in English for code. Editing does not run a model or change the authored answers above.
REQUEST.JSON
Python 3.10+; no third-party dependencies. Run from the directory containing examples/.
Download all example codepython -m examples.hello_jev reviews --mode offline --language enThis checks the local workflow, not model quality. Changed inputs cannot replay old fixture answers.
Set TYPESAFE_API_KEY locally and put request.json in the project root. Live calls send material to TypeSafe and may incur charges. Live interoperability has not been tested in this edition.
python -m examples.hello_jev reviews --mode live --language en --request request.jsonOpen the official Playground · Account and environment setup
Request fields checked against the official API reference on 2026-09-20. API reference