Tell Evoduce the rules you already know. They answer first; Jev handles the decisions they cannot settle. Confident answers can become reusable rules you review and approve. The same learning layer works with other decision models, while an NS vocabulary keeps typed choices consistent.
Review the pattern, then let your rule answer next time.
Evoduce uses the same typed question and customer rules whether the next answer comes from Jev, a configured OpenAI-compatible model, or the Decisions API.
Define what must happen for cases you already understand. A matching rule can settle a question before any model call.
Route unresolved questions, reuse previous answers, and propose candidate rules from confident repeated results. You control which rules are used.
Use Jev, a configured OpenAI-compatible model, or rules-only mode through the Evoduce Decisions API.
Use verified vocabulary terms as typed Choice options. Structural checks govern additions to that vocabulary; a model's choice remains a judgment, not a proof of the label.
Jev and configured OpenAI-compatible models work with the Decisions API today. An AWS Automated Reasoning connection is planned for policy checks and findings.
Start with Jev's typed Choice API and put your own rules and learning around it.
Configure a chat-completions endpoint for typed decisions with the same rules-first path.
Send typed questions directly. Use your rules, a configured provider, or both.
Planned: Evoduce for AWS Automated Reasoning, connecting policy findings to the same review workflow.
Evoduce