/ 01
The Core Difference
Jev receives state plus typed questions and returns structured decisions. An LLM primarily generates text. That difference changes how applications validate, parse and act on the output.
/ 02
Side-by-Side
| Need | Use Jev when… | Use an LLM when… |
|---|---|---|
| Output | A named option, ordered score or yes/no probability | You need prose, code or open-ended content |
| Control | Your code defines valid options and thresholds | The answer space cannot be fully bounded |
| Uncertainty | You can use probabilities and Choice/Score confidence | You will evaluate quality by another method |
/ 03
Compose Both
Use Jev to route or verify an LLM workflow, and use an LLM when the workflow truly needs generation. Keep arithmetic and invariant checks in code. Explore Noul, JSON output and the Jev playground.
/ FAQ
Frequently asked questions
Is Jev a replacement for an LLM?
No. They fit different output needs; Jev does not generate prose.
Can Jev return free-form text?
No. Its documented answer types are Choice, Score and Noul.