Metadata-Version: 2.4
Name: terrarium-treehouse
Version: 0.1.1
Summary: Treehouse is an external observability and interpretability tool for behavior-tree-driven agents
Project-URL: Homepage, https://github.com/ricardokirkner/terrarium
Project-URL: Documentation, https://github.com/ricardokirkner/terrarium/tree/main/docs
Project-URL: Repository, https://github.com/ricardokirkner/terrarium.git
Project-URL: Issues, https://github.com/ricardokirkner/terrarium/issues
Author-email: Ricardo Kirkner <ricardo@kirkner.com.ar>
License: Apache-2.0
Requires-Python: >=3.13
Requires-Dist: terrarium-vivarium
Provides-Extra: visualizer
Requires-Dist: fastapi>=0.115.0; extra == 'visualizer'
Requires-Dist: uvicorn>=0.34.0; extra == 'visualizer'
Requires-Dist: websockets>=15.0; extra == 'visualizer'
Description-Content-Type: text/markdown

# 🌳 Treehouse

Treehouse is an **external observability and interpretability tool** for behavior-tree–driven agents.

It lets you observe how behavior unfolds in real time — which branches were evaluated, which paths were taken, and where execution stalled or failed — without interfering with execution.

Treehouse is where you go to *watch behavior trees think*.

---

## What Treehouse Does

- Consumes semantic execution events
- Reconstructs behavior tree execution over time
- Surfaces decisions in structural, human-readable form

Treehouse treats behavior trees — not logs or metrics — as the unit of meaning.

---

## Event Boundary

Treehouse relies on the shared event boundary defined here:

> `../docs/event-boundary.md`

This boundary is Treehouse’s primary input API. Any runtime that emits compatible events can be observed — Vivarium is a natural fit, not a requirement.

---

## What Treehouse Is *Not*

- Not a runtime
- Not a controller
- Not coupled to a specific agent framework

Treehouse observes. It does not direct.

---

## Status

Treehouse is early-stage and exploratory.

The focus is on correctness, clarity, and faithful reconstruction of behavior — polish comes later.

---

*Run the behavior. Observe the decisions. Understand the agent.*
