NeutronOS
Nuclear Energy Unified Technology for Research, Operations & Networks
A modular digital platform for nuclear facilities that unifies model management, operations tracking, signal analysis, and knowledge retrieval — replacing fragmented workflows (paper logs, spreadsheets, phone calls) with integrated digital tools.
Built on top of Axiom (domain-agnostic platform framework). NeutronOS adds nuclear-specific knowledge, agents, facility packs, and tools.
Install
pip install neutron-os
That installs the whole platform: the neut CLI, the Model Corral, Materials
Database, Facility Packs, the Experiment Manager, and the Axiom-provided chat /
knowledge-retrieval / signal / publishing surfaces. No physics code
required — operations tracking, experiment scheduling, the model registry,
and knowledge retrieval all work on their own.
Verify:
neut --help # NeutronOS CLI
axi doctor # environment + dependency check (Axiom-provided)
NeutronOS bundles axiom-os-lm (the Axiom platform) and adds the nuclear-domain layer on top.
Physics codes (optional)
NeutronOS runs transport/physics codes through optional, pluggable adapters —
per ADR-018 and ADR-044, each code is its own AEOS-conformant extension, shipped as a pip extra you install only when you run that code. NeutronOS is not tied to any one code; MCNP, MPACT, Serpent and others install the same way (neutron-os[mcnp], …) as their adapters ship.
OpenMC is the reference adapter. Install its extra (the wrapper) plus the native code:
pip install "neutron-os[openmc]" # the Axiom OpenMC adapter (the extra)
conda install -c conda-forge openmc # the native physics code; full Python API
Then both surfaces work — and the wrapper is non-shadowing, so plain OpenMC is untouched:
| Surface | Command | Notes |
|---|---|---|
| Traditional OpenMC | openmc -i geometry.xml, import openmc; openmc.run() |
Exactly as the OpenMC docs describe |
| NeutronOS-routed | neut model run pin-cell --on local:openmc --tail |
Adds a live convergence dashboard, signed receipts, federation routing |
Prefer Docker to a local OpenMC build? The adapter has a Docker runner (kernel_options={"runner": "docker"} pulls openmc/openmc:latest) — skip the conda step entirely.
From source (developer install):
git clone https://rsicc-gitlab.tacc.utexas.edu/ut-computational-ne/neutron-os-core.git
cd neutron-os-core
pip install -e ".[all]"
# Or use the bootstrap script (creates venv, installs, sets up direnv)
./scripts/bootstrap.sh
See CONTRIBUTING.md for detailed setup. The first-day install walkthrough with troubleshooting lives at docs/guides/first-day-openmc-install.md.
What NeutronOS Does
NeutronOS ships nuclear-domain builtins; the AXI orchestrator, chat, signal, and publishing surfaces are provided by the underlying Axiom platform.
Researcher / Operator
|
v
neut model ──> Physics model registry (Model Corral)
| init, validate, add, search, generate, review, share...
|
neut facility ──> Facility packs (NETL-TRIGA, MSRE, PWR-generic)
| Materials database, configurations, procedures
|
neut experiment ──> Experiment Manager (neut exp) — sample lifecycle
|
neut demo ──> Guided walkthroughs
+ Axiom-provided surfaces (chat assistant, signal pipeline, document
publishing) layered on top via the Axiom platform.
Key Capabilities
Model Corral (24 CLI commands)
Physics model registry with auto-detection, validation, material generation, and collaboration:
neut model add ./input.i # Auto-detect code type (MCNP, MPACT, etc.)
neut model validate ./input.i # Validate model file
neut model search --code mcnp # Search by code type
neut model list # Browse registered models
neut model materials --card UO2 # Generate MCNP/MPACT material cards
neut model materials # List verified materials (11 compositions)
neut model review <id> # Start collaborative review
neut model share <id> # Share via federation
All commands support --json. Full list: init, validate, add, clone, search, list, show, pull, lineage, diff, export, audit, generate, lint, sweep, materials, share, receive, review, reviews, resolve, invite, contributors, status.
Materials Database
11 verified compositions from authoritative references (PNNL-15870, NUREG/CR-6698, and facility-measured characterizations cited by page):
neut model materials # List all materials
neut model materials --card UO2 # Generate MCNP cards
neut model materials --card Zircaloy-4 --format mpact # MPACT format
MaterialSource protocol with 5 sources and priority merging. Deterministic MCNP/MPACT card generation.
Facility Packs
Pre-built knowledge bundles for specific reactor types:
neut facility list # Available packs
neut facility install NETL-TRIGA # Install a facility pack
neut facility show NETL-TRIGA # Pack details
neut facility materials NETL-TRIGA # Facility-specific materials
neut facility sync # Sync with upstream
Shipped packs: NETL-TRIGA, MSRE, PWR-generic. Each includes materials, configurations, and domain knowledge.
Interactive Chat (Axiom-provided)
Chat is provided by Axiom's AXI orchestrator, not a NeutronOS builtin. NeutronOS contributes the nuclear knowledge, materials database, and facility grounding.
Nuclear engineering assistant with tool calling and RAG grounding:
axi chat # Start interactive session
axi chat --resume # Continue last session
axi chat --simple # Read-only mode (no writes)
The assistant has access to the materials database, model registry, and facility knowledge. When asked about material compositions, it uses verified database values — never training data.
Signal Pipeline (Axiom-provided)
Signal ingestion is provided by Axiom's SCAN agent; there is no
neut signalnoun registered in NeutronOS today.
Proactive program awareness — ingest signals from multiple sources:
axi signal brief # Catch up on what happened
axi signal draft # Generate weekly changelog
axi signal status # Pipeline health
axi signal watch # Watch for new signals
Sources: voice memos, Microsoft Teams, GitLab, freetext. Extracts: decisions, action items, risks, blockers, deadlines, people mentions, technical concepts.
Document Publishing (Axiom-provided)
Document publishing is provided by Axiom's PRESS agent; there is no
neut pubnoun registered in NeutronOS today.
Markdown to DOCX lifecycle with review gates:
axi pub overview # Document ecosystem dashboard
axi pub publish report.md # Generate + publish
axi pub review report.md # Interactive review
axi pub status # Document status
CoreForge Bridge
Integration with CoreForge neutronics tools for MCNP model workflows.
Platform Features (via Axiom)
NeutronOS inherits all Axiom platform capabilities. These ship with the Axiom
platform and are invoked as axi <noun> (not registered NeutronOS nouns):
| Feature | Command | Description |
|---|---|---|
| Federation | axi federation |
Node identity, peer discovery, trust, .axiompack distribution |
| RAG | axi rag |
Three-tier knowledge corpus (community / org / personal) |
| Knowledge | axi knowledge |
Observatory — velocity, accumulation, impact metrics |
| Research | axi research |
Call to Research — distributed research coordination |
| Security | axi security |
TRIAGE — content verification, anomaly detection |
| Connections | axi connect |
Manage external service credentials |
| Diagnostics | axi doctor |
TRIAGE — AI-powered environment health checks |
Repository Structure
Neutron_OS/
├── src/neutron_os/ # Python package root
│ ├── cli.py # CLI entry point
│ └── extensions/builtins/ # Builtin extensions
│ ├── model_corral/ # Model Corral (neut model, neut facility)
│ ├── expman/ # Experiment Manager (neut experiment / neut exp)
│ └── demo/ # Guided walkthroughs (neut demo)
├── webapp/ # Web experience
├── runtime/ # Instance data (config, inbox, sessions)
├── tests/ # Cross-cutting tests
├── docs/ # Documentation — see docs/README.md
│ ├── prds/ specs/ adrs/ # One folder per doc kind
│ └── guides/ runbooks/ reference/ strategy/ templates/
├── infra/ # Deployment configs (K3D, Docker)
├── scripts/ # Bootstrap and install scripts
└── pyproject.toml # Package definition
The agent fleet (AXI/SCAN/TIDY/RIVET/TRIAGE/PRESS/CURIO/…), gateway/orchestrator, and shared infra helpers (
axiom.infra.*) are provided by Axiom, not this repo.
Development
# Run all tests (1700+)
make test
# Run specific suites
pytest tests/ -v # Cross-cutting
pytest src/neutron_os/extensions/builtins/model_corral/tests/ -v # Model Corral
pytest -m "not integration" # Unit only (runs as pre-push hook)
# Lint (must pass before push)
ruff check src/ tests/
ruff format src/ tests/
# Build
make build
Tech Stack
| Layer | Technology |
|---|---|
| CLI & Agents | Python (argparse, rich, prompt-toolkit) |
| LLM Gateway | Model-agnostic (Anthropic, OpenAI, Qwen, any OpenAI-compatible) |
| Database | PostgreSQL + pgvector |
| RAG | pgvector embeddings, hybrid vector+full-text search |
| Infrastructure | K3D (local), Docker Compose (fallback) |
| CI/CD | GitLab CI |
| Packaging | Hatchling (pip install neutron-os) |
| Platform | Axiom (axiom-os-lm) |
Design Principles
- Reactor-agnostic core with facility-specific configuration via packs
- Offline-first — queue locally, sync on restore
- Human-in-the-loop for all writes in safety-adjacent contexts
- No vendor lock-in — model-agnostic, cloud-agnostic, IDE-agnostic
- Provider pattern — swap implementations via config, not code changes
- Federation-native — every node is sovereign; invite = join as peer, not shared DB user
Documentation
docs/ holds one folder per document kind — start at the map in
docs/README.md:
- Executive PRD — what we're building and why
- Master Tech Spec — how it's built
- ADR index — why it's this way (decision records)
- Glossary — what the words mean
- New PRD / spec / ADR? Start from docs/templates/; the
portfolio-wide standard lives in the Axiom repo
(
docs/conventions/doc-standards.md) and is audited byaxi hygiene stat docs
Releases
Releases ship from tags to PyPI; the version badge above tracks the latest. See CHANGELOG.md for full release notes.
Contributing
See CONTRIBUTING.md for development workflow, git practices, and conventions.
License
Apache 2.0 — see LICENSE.
Acknowledgments
Developed at The University of Texas at Austin, Department of Mechanical Engineering — Nuclear & Radiation Engineering Program.
Metadata
Release files for neutron-os 1.15.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| neutron_os-1.15.1.tar.gz | 1.2 MB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| neutron_os-1.15.1-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 2.8 MB
Release files / neutron_os-1.15.1.tar.gz
| Download URL | neutron_os-1.15.1.tar.gz |
|---|---|
| Size | 1.2 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
f157b19d6e4d576f75aa0c77979b006d9e174c6639e9b216b02c99371209203d
|
|
BLAKE2b-256 checksum How to use checksums |
d208e8cc8e7f8d63ff0c870e0f0d2248982f46239153e97be22f17a0a41b0ae6
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 6, 2026.
Transparency logRelease files / neutron_os-1.15.1-py3-none-any.whl
| Download URL | neutron_os-1.15.1-py3-none-any.whl |
|---|---|
| Size | 1.6 MB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
19361f40cae191efd43f97c1a9201f2e941f6e82821136cd3c337d0b49e60661
|
|
BLAKE2b-256 checksum How to use checksums |
e3089ba1bf3677a80cb66fa51d82dfe1583a29ed6c0c6e97b37c60e59e67d703
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Oct 6, 2026.
Transparency log