What it does
The middleware consumes Choice answers to select an agent model and Noul probabilities to block configured risky tools. Version 0.0.1a3 is explicitly a pre-release.
Integrations
langchain-typesafe
Adds Jev's typed decisions as a LangChain Runnable, with experimental middleware for model routing and tool-risk checks.
Screenshot unavailable. Open the experiment ↗
How you can use it
Your developer can use this tool to examine a user's message. They send the text to TypeSafe, an outside AI service. The app can then ask specific questions, like whether a support request is urgent. Your developer needs an access key to connect the app.
The software also includes early tests for checking risky actions. The app can review what a program is about to do based on your instructions. It blocks an action marked dangerous before it happens. Because this is an early version, these features might change without warning.