ABRIA overview
What ABRIA 0.24.0 is, how ABR differs from a foundation model, and how the native C++ and Python layers fit together.
Source map
- python/abria_server/__init__.py
- python/abria_server/api.py
- python/abria_server/abr_runtime.py
- cpp/include/abria/abr.hpp
ABRIA 0.24.0 in one view
The uploaded runtime identifies itself as ABRIA 0.24.0. The Python service exposes the public API and orchestrates native C++ cognition, per-user ABR fabric, memory, tooling, workspaces, research and optional learning layers.
In ABR mode, the primary runtime does not require a pretrained foundation checkpoint. The native AbrBrain grows a user-specific cognitive fabric from experiences: concepts, relations, constructions and reusable relation-composition programs.
- Python runtime version: 0.24.0
- Native version is reported separately by /api/health
- Per-user ABR state is isolated by user identity
- Guest sessions are isolated but non-persistent
Two intelligence paths
ABRIA contains more than one execution path. ABR mode is the personal cognitive-fabric runtime; the broader codebase still contains optional neural/foundation components used by other modes and experiments.
The docs call this distinction out explicitly so 'ABRIA has neural code' is not confused with 'ABR itself is a neural network'.
- ABR: concepts, relations, constructions, programs, theories/abstractions and personal fabric
- Optional neural_v2/foundation/online-learning modules remain separate layers
What the docs cover
These pages document the architecture from the uploaded C++ and Python trees, plus the modern ABRIA workspace UI pattern already used by the project.
- C++ native brain and cognitive fabric
- Python API, identity and AETHER bridge
- Workspace/file/code intelligence
- PABR personal twin and media
- Safety, research, tools and verification