A global node vault for AKMS: 264 typed knowledge nodes covering computational mechanics, FFT-Galerkin micromechanics, constitutive modelling, phase-field fracture, solvers and preconditioning, and the MOOSE framework.
This repository is content, not code. It ships no Python package. AKMS compiles it into a directed graph and projects task-scoped slices of it into agent context.
pip install akms
akms vault install https://github.com/CEmM2/akms-vault-compmech/archive/refs/tags/v1.0.0.tar.gz
akms vault statusvault install also accepts a local directory or a .tar.gz, so a clone works
just as well:
git clone https://github.com/CEmM2/akms-vault-compmech
akms vault install ./akms-vault-compmechIt copies only vault content — nodes and their content/ payloads. This
README, the licence and the CI configuration are left behind deliberately: the
AKMS graph compiler treats any markdown in a vault that lacks akms_schema as
a schema error, so a README copied into the vault would break every build.
Nothing is installed automatically and nothing is overwritten. vault install
refuses to replace a vault that already holds nodes unless you pass --force.
| Domain | Nodes |
|---|---|
| computational-mechanics | 160 |
| fft-galerkin | 30 |
| constitutive | 26 |
| phase-field | 22 |
| architecture (MOOSE) | 14 |
| solver | 8 |
| gpu-simulation | 4 |
| total | 264 |
Nodes are organised by subject directory. skills/ holds fourteen
cross-cutting nodes — a notation cheatsheet, a verification-benchmark index,
and topic hubs the MOOSE nodes link into — and content/ holds the payload
bodies those nodes address through content_ref.
Nodes are original synthesis: equations, algorithms and methods restated in the authors' own words, with the source literature cited in the body. They are not reproductions of any source text.
The 70 nodes under frameworks/ are derived from
DeepWiki's index of idaholab/moose and
each carries a provenance note saying so. MOOSE is developed at Idaho National
Laboratory and licensed LGPL-2.1. Those nodes are derived summaries, not MOOSE
documentation — consult upstream before relying on a detail.
Every node is Markdown with YAML frontmatter conforming to AKMS schema v2:
---
akms_schema: v2
id: fft-green-operator
title: Green Operator for the Lippmann-Schwinger Equation
domain: fft-galerkin
status: established
confidence: 0.9
source: hybrid
edges:
- to: fft-lippmann-schwinger
type: requires
weight: 0.8
---edges name other node ids. Every edge target in this vault resolves within
this vault — CI enforces that, so a release can never ship a broken reference.
A vault improves by being corrected. Open a pull request; CI validates that every node parses as schema v2, that ids are unique, and that no edge points at a node that does not exist.
Run the same checks locally:
pip install akms pyyaml
python scripts/validate_vault.pyReleases are tagged vMAJOR.MINOR.PATCH and are independent of the akms
package version — nodes change far more often than code does. Each tag has a
GitHub-generated tarball, which is what akms vault install fetches.
- major — nodes removed or ids renamed
- minor — nodes added
- patch — corrections that leave the id set unchanged
Released under CC0 1.0 Universal — public domain dedication. You may use, adapt and redistribute this material for any purpose, commercial or not, without permission or attribution.
Use is entirely at your own risk. CC0 §4(b) offers the work as-is with no representations or warranties of any kind, and §4(c) disclaims responsibility for clearing rights of other persons that may apply. Nothing here is verified for correctness, fitness for any purpose, or freedom from error. It is a research aid, not a reference: check any result that matters against the primary literature.
Attribution is not required, but if this material is useful in published work, a citation is appreciated — see CITATION.cff.