Virtuallaborer/tesla-core-protocol ? reverse-engineered prompt

Build me a Python library for a deterministic agent reasoning system.

I want it to model observations, provenance, and reasoning trees in a way that always gives the same results for the same inputs. The core should keep timestamps, IDs, confidence, and source history consistent, and it should enforce those rules with tests. It should also support a deterministic reasoning flow that can combine multiple steps, prune branches, and keep tree identity stable.

Please include the later identity layers too, like semantic and epistemic identity, temporal identity, agent identity, execution loop tracking, and the temporal field surfaces. The behavior should feel strict and predictable, with clear hashes and continuity across sessions and episodes.

Use the existing project structure, keep the implementation clean and minimal, and make sure the test suite passes. If you need to check current Python docs or best practices while working, feel free to look them up online.