Jev in production › SDKs and workflow integrations
Vybez
TypeScript library whose is, pick and rate tagged templates embed semantic rules in ordinary conditionals, answered by Jev by default as probabilities, choices or ratings; questions on the same state in one tick share one request.
- Use
- SDKs and workflow integrations
- Industry
- Developer tools
- Form
- Open-source tool
- Stage
- In production
- Listed
- 2026-09-24
- Found via
- github
- Repository
- clavia-labs/vybez
- Stars
- 7
- Forks
- 0
- Last push
- 2026-09-24
- Language
- TypeScript
- License
- MIT
The README opens with
Vybez embeds semantic rules in code, weaving model judgments with exact logic through ordinary conditionals and functions. Questions return probabilities, choices, or ordered ratings using Jev by default. Typed refs connect questions to state, TypeScript validates refs and choices, and questions in the same tick are automatically batched.
- Soft and hard logic together. Ask a model where rules stop being useful, then branch with normal TypeScript. - Answers you can compute with. Probabilities, choices, and ratings are data, not text that application code must parse. - Provider neutral. Jev is the default, with prompt models and custom providers available through the same state API. - Batched by state. Independent questions asked together share...
Badge
For the project's own README, linking back here:
Also used for sdks and workflow integrations
- JarvisCore (Prescott Data) 434 ms median warm
- Jev for Apple Foundation Models 40–150ms
- Laravel Jev 70–500ms
- JevForce 11.1×
- Semantic Validator 90
- openpave-jev 35 to 500 milliseconds
- Yao Agents 70–500ms
- Juspay NeuroLink ~$0.00002/decision