Jev examples for
decisions inside software.
Jev is TypeSafe AI’s System One model. It does not write a reply. It evaluates state against typed questions and returns bounded decisions with probabilities—so ordinary code can decide what happens next.
State in. Decisions out.
Give Jev the relevant state and a narrow question with a defined answer space. It returns values your application can validate, threshold and route.
- 01Describe the state
Include only the context needed for the decision.
- 02Define bounded questions
Ask for choices, scores or yes/no probabilities.
- 03Let code own the action
Enforce permissions, thresholds and fallback paths outside the model.
Start with one small decision.
Agent Tool Selection
Choose the next tool or specialist agent from a bounded set, then let ordinary code validate and execute it.
Open pattern ↗02AI Model Routing
Send routine decisions to a low-cost path and reserve slower reasoning models for genuinely difficult requests.
Open pattern ↗03RAG Passage Filtering
Score retrieved passages for relevance, answer support, contradiction and prompt-injection risk before generation.
Open pattern ↗04Prompt-Injection Detection
Inspect retrieved or user-supplied content before an agent lets that content influence tool use.
Open pattern ↗05Citation Verification
Compare a claim, its cited passage and source context to classify whether the citation supports the claim.
Open pattern ↗06Browser Automation
Turn the current page into a closed set of possible actions and ask Jev which action best advances the goal.
Open pattern ↗