typesafe-decisions¶

Small, typed building blocks for semantic decisions in Python.

typesafe-decisions is an experimental convenience layer around TypeSafe System One. It turns semantic judgments into ordinary Python values that an application can use for routing, thresholds, and other decisions.

Get started API reference

Experimental and unofficial

This project is not an official TypeSafe SDK. It is pre-release software, and its public API may change.

Choose the right operation¶

Operation	Use it when you need	Result
`DecisionClient.feels()`	A semantic score for one property	`float`
`DecisionClient.match()`	The best label from explicit criteria	`str` or a Pydantic model
`llm()`	Generated text through Cloudflare AI Gateway	`str`

The first two operations use TypeSafe System One. The standalone llm() helper is independent and sends only its instruction and explicitly supplied context to Cloudflare.

Small public interface¶

from typesafe_decisions import DecisionClient, llm

The package exports these two names deliberately. Configuration and advanced request controls remain available without expanding the common call path.

Next steps¶

Follow Getting started for a first routing decision.
Adapt the complete application examples for support routing, escalation, and response drafting.
Read Concepts to understand the boundary between decisions and text generation.
Use the task-focused guides for concrete patterns.