Agentic exokernel for governed multi-system AI runtimes
Project description
AAAX
Agentic exokernel for governed multi-system AI runtimes.
AAAX sits above sssn and lllm and gives that stack an operating-system-style control plane:
- dock and wire modules into a governed constellation
- issue and revoke capabilities for mediated resources
- authorize side effects through an action gate
- manage lifecycle events like revoke, pause, resume, and drain
- bridge the default LibOS to
lllm - provide the first-class kernel contract for higher applications such as the Productive Suite
The current codebase is the first kernel slice: bootstrap, governance channels, policy engine, module loading, lifecycle control, and a minimal LLLM bridge are implemented and tested.
Install
pip install aaax
AAAX targets Python 3.10+ and installs its framework dependencies through PyPI:
sssnfor the network substratelllm-corefor the default LibOS bridge
Quick Start
Create a minimal kernel config:
[aaax]
id = "aaax-main"
name = "AAAX Kernel"
policy = "default"
[aaax.libos]
name = "lllm"
strict_boot = true
discover_shared_packages = false
[aaax.network]
publish = false
host = "0.0.0.0"
port = 8100
[[aaax.modules]]
id = "example-agent"
framework = "custom"
channels = []
executors = []
remote_channels = []
Launch the kernel:
aaax launch aaax.toml
Or from Python:
import asyncio
from aaax import AAAXConfig, bootstrap_kernel
async def main() -> None:
config = AAAXConfig.from_file("aaax.toml")
kernel = await bootstrap_kernel(config, start_channels=True)
await kernel.step()
asyncio.run(main())
Architecture Stack
Productive Suite and other AAAX applications
|
AAAX
|
LLLM (default LibOS) + other LibOS adapters
|
SSSN
SSSNremains the transport and topology substrate.LLLMremains the first LibOS and package/runtime layer.AAAXowns governance, module trust, capability issuance, action authorization, and lifecycle.Productive Suiteis the first citizen of AAAX, not a kernel dependency.
Project Layout
aaax/ Python package
docs/ MkDocs site for aaax.one
notes/ design and implementation notes
sssn/ vendored local framework source for development
lllm/ vendored local framework source for development
Documentation
Full documentation is published at aaax.one.
- Getting started: installation, config, and first launch
- Architecture: kernel functions, framework boundaries, Productive Suite positioning
- Core reference: config, governance, module loading, and LibOS bridge
- API reference: public AAAX Python entry points
Development
For monorepo development against the local framework copies:
pip install -e ./sssn
pip install -e ./lllm
pip install --no-deps -e .
pip install -r requirements-monorepo.txt
pytest aaax/tests -q
python -m mkdocs build --strict
python -m build
The published package still works outside this repo through normal PyPI dependencies. Inside the monorepo, _vendor.py prefers the checked-out sssn/ and lllm/ trees when they exist.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file aaax-0.1.0.tar.gz.
File metadata
- Download URL: aaax-0.1.0.tar.gz
- Upload date:
- Size: 31.5 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.14.3
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
f790455bb797a3f7448fa71a27865b81e4e99278dab59a030ade45c951eb7747
|
|
| MD5 |
fb6ebfd33c652495f6bf10984c7e2e1a
|
|
| BLAKE2b-256 |
816561c298d0574f048eb11a6a9e8a70e3ae1a036b520e0f5666e02a6be0fe99
|
File details
Details for the file aaax-0.1.0-py3-none-any.whl.
File metadata
- Download URL: aaax-0.1.0-py3-none-any.whl
- Upload date:
- Size: 23.7 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.14.3
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
308d26c1ef43baf525dc567fe50c348892e74726ecf1a920e8c7a195b27a156c
|
|
| MD5 |
71c5052deb53a4117e5a9c470a6438b6
|
|
| BLAKE2b-256 |
dca8f2c64f93822a5afbcd1dab9c765d93f2bc81d547623ce73e5ad6f5af0734
|