Your first decision with Jev
Define a question, evaluate an example and find the official integration guide.
1. Identify the decision
Pick a task with outcomes you can describe and check. Routing a request, scoring urgency or checking whether a condition holds are useful starting points. Record how the decision is made today so you have something to compare against.
2. Choose a primitive
- Choice: select from a defined set of options.
- Score: evaluate a state against a supplied rubric.
- Noul: assess the truth of a statement.
- Box (experimental, Jev 1.14): locate one image target. It requires the images opt-in header and a separate presence check; see the Images and Box checklist.
Our local explorer illustrates these concepts without calling the model. Its values are teaching examples, not actual inference results.
3. Use the official service
Visit TypeSafe AI to request early access. Once access is available, follow the official quickstart for the current Playground, API authentication and supported request format. A jevai.me account does not provide model credentials.
4. Evaluate before connecting actions
Start with a small set of representative inputs and expected decisions. Include ambiguous requests and incomplete context. Review mistakes before using a result to trigger an external action.
Keep credentials on your server. Apply your application’s normal authorization checks to any consequential action, regardless of the model output. For a fuller rollout process, see the building path.