[[ABIO]]
Alien Biology — Documentation¶
A framework for testing agentic AI reasoning through procedurally generated biological systems untainted by training data. Alien Biology measures complex, agentic reasoning/learning that is:
- REAL-WORLD — practical, complex, real-world-relevant agentic reasoning tasks.
- UNTAINTED — draws its tests from an "Alien" universe, avoiding confounds with LLM training corpora.
- CONTROLLABLE — parametrically constructed for fine-grained analysis of the limits of agentic reasoning.
How this documentation is organized¶
This tree is mirrored between the code repository (docs/) and the vault (ABIO Docs/) — the folders and paths are identical on both sides. Hand-authored folders sync both directions; generated folders flow one way from the code.
| Folder | Contents | Sync |
|---|---|---|
| [[ABIO Architecture Docs|Architecture/]] | System design — subsystem overviews, PRDs, spec language, per-class and per-command reference. | bidirectional |
| [[ABIO Alienbio User Guide|Guide/]] | How to use the system — core specs, generators, execution, the agent interface. | bidirectional |
| Modules/ | Auto-generated Python API reference (from docstrings). | one-way (code → vault) |
| diagrams/ | Diagram assets (svg/dot). | one-way (code → vault) |
Start here¶
- ABIO Architecture Docs — the architecture index: subsystems, data model, protocols, and the benchmark-generation design (ABIO Suite Construction, Scenario Generator PRD).
- ABIO Alienbio User Guide — usage tutorials: ABIO Expr Spec, ABIO Expr Python API.
- API Reference — auto-generated class and function docs.