JevModel is a structured decision service built on TypeSafe AI's typesafe/jev-1.13 model. Instead of generating free-form text, it accepts a state and a set of focused questions, then returns typed answers your application can check and act on. Each answer comes with a probability, so code can branch on confidence rather than guess at prose.
How decisions work
The model supports three answer types. Choice selects one of your supplied options, such as a support queue or moderation label. Score assigns a discrete level from a scale you define, such as lead priority. Noul makes a yes/no judgment against a criterion, such as whether a case needs review. Because the output is constrained to your options, it is straightforward to validate and route in production code.
Building and testing flows
Decision flows let you chain questions on a canvas: a spam check, then intent, then routing and urgency. Each answer leads to the next question or to an outcome, and a single API request runs the whole flow while only the questions on the selected path are billed. Templates for support inbox triage, comment moderation, and lead qualification provide a starting point that can be edited freely. Before shipping, you can paste real texts and watch the path light up with the probability of every answer, and borderline results can take their own branch so a person sees what the model is unsure about.
Batch evaluation
Batch processing accepts CSV, TXT, or JSON Lines files up to 1,000 rows. You can run a template or your own question over the data and inspect every structured answer, then export the results as CSV. This makes it practical to compare wordings, review results on your own texts, and choose a threshold before production.
API and console
A single API endpoint handles Choice, Score, and Noul requests with typed answers. API keys are created in the Console, which also manages saved flows, recent runs, and token balance. The playground, batch jobs, and API share one account balance, and failed requests are not charged.
Common applications
JevModel fits workflows that end in a decision: content moderation, spam and scam filtering, support triage, ticket routing, priority and escalation, language detection, personal data screening, prompt injection screening, and sentiment tracking. The service is independent of TypeSafe AI and calls the model through OpenRouter. A free tier provides three no-signup runs per day and 100,000 free tokens after email verification, with token packs available for ongoing use.
