Jev in Practice

Practical examples of using Jev (TypeSafe AI's decision model) in real applications, built with LangChain.

An LLM writes. Jev decides. You give Jev your options and it picks one, tells you how sure it is, and never answers outside your list.

What's inside

Example	What Jev decides
Support ticket triage	Which team, how urgent, whether the customer wants a refund
Model routing	Small, premium, or private model for each request
Reply guardrails	Whether an AI-written reply is safe to send
Tool selection	Which tool an agent should use, and blocking risky tool calls
Finance inbox	Document type, urgency, and "bank details changed" fraud

• jev_decision_guide.ipynb: the notebook, with outputs from a real run

Setup

1. Install the dependencies:

   pip install "langchain-typesafe[experimental]" langchain-groq python-dotenv pandas

2. Create a .env file with your keys (Jev runs through OpenRouter, the LLMs through Groq):

   OPENROUTER_API_KEY=your_openrouter_key
   GROQ_API_KEY=your_groq_key
   

3. Open jev_decision_guide.ipynb and run it top to bottom.

Running in Colab? There's no .env file there. Add both keys under Secrets (the key icon in the left sidebar), then run this before the setup cell:

import os
from google.colab import userdata
os.environ["OPENROUTER_API_KEY"] = userdata.get("OPENROUTER_API_KEY")
os.environ["GROQ_API_KEY"] = userdata.get("GROQ_API_KEY")

Note: langchain-typesafe is in early alpha and its middleware is experimental, so the API may change.