Jev TypeSafe AI

Jev TypeSafe AI is a small tutorial project demonstrating how to use the TypeSafe AI SDK to classify a user's response and choose an appropriate reply. The interactive lost-and-found example asks whether the user lost something, evaluates the answer, and routes the response based on the lost_something score:

• Above 0.8: directs the user to the lost-and-found counter.
• Below 0.2: asks what else the user needs help with.
• From 0.2 through 0.8, inclusive: asks the user to clarify.

Requirements

• Python 3.13 or newer
• uv for managing the project and its dependencies
• TypeSafe API configuration for the interactive example

The project dependencies are declared in pyproject.toml and installed by uv:

• typesafe-sdk for the TypeSafe client and decision model
• python-dotenv (imported as dotenv) to load configuration from a .env file

Setup

From the project root, install the dependencies:

uv sync

Create a .env file in the project root with your TypeSafe configuration:

TYPESAFE_API_KEY=your_api_key
TYPESAFE_DEFAULT_MODEL=your_model
TYPESAFE_BASE_URL=your_base_url

Use the model name and base URL for your TypeSafe service. Keep the API key private and do not commit the .env file.

Run

Run the interactive lost-and-found example:

uv run python src/jev_typesafe_ai/lost_found.py

Enter a response when prompted. Type exit to quit; press Ctrl+C to exit as well.

The top-level main.py is a separate starter stub that prints a greeting; it does not launch the interactive example. Run it with:

uv run python main.py