Skip to main content

Third Matter Specification — declarative YAML metalanguage and toolchain for DFT/MLIP/MD calculations with sanity gates.

Project description

TM-Spec

Third Matter Specification — declarative YAML/JSONL metalanguage and reference toolchain for DFT/MLIP/MD calculations on atomistic systems, with sanity gates, provenance, NOMAD import/export, and code-agnostic level-of-theory.

Status: v0.3 (draft) · Package: 0.3.0 · License: MIT (code) + CC-BY-4.0 (spec) · Spec home: https://exopoiesis.github.io/tm-spec/


What is it

One YAML file describes a structure, defects, magnetic state, environment, reaction network, DFT/MLIP calculation, workflow (NEB / US / MetaD / MD / MLIP benchmark / single point / relaxation), results, sanity gates (post-hoc) and an optional pre-flight block (predictive), and provenance in a form that:

  • Reviewers can read in any text editor without runtime.
  • Validators can check against JSON Schema 2020-12 (schemas/0.3.json; legacy 0.2 and 0.1 documents remain supported).
  • Tools can ingest: extract from existing ASE/QE/CP2K/ABACUS Python scripts; import public NOMAD archive entries; export TM-Spec documents as a NOMAD upload bundle; lint pilots vs paired scripts; auto-fill sanity gates from run artefacts.
  • Cross-codes: the same file describes Quantum ESPRESSO, CP2K, ABACUS, GPAW, MACE, CHGNet, or an imported NOMAD calculation at the same level of detail.

Why

Paper SI today is often either a code-specific input dump (pwscf.in) or a workflow archive that requires specialized runtime to inspect. TM-Spec is the missing middle: one human-readable YAML file per calculation, validatable, exportable to FAIR archives, and reproducible via paired script lint. Zero runtime dependencies for inspection.

What is not the goal

  • Replace AiiDA/atomate2 workflow management. They own provenance graphs during work; TM-Spec owns the paper-grade snapshot.
  • Be Turing-complete. YAML records declarative state and a pointer to a paired script or external archive entry.

Quickstart

pip install tm-spec       # first PyPI release: v0.2.0

# validate one file
tm-spec validate examples/pyr_smoke.tm.yaml

# validate all bundled examples
tm-spec validate --all examples --strict

# extract a stub from an ASE/QE/CP2K script
tm-spec extract path/to/neb_canonical.py --out tmp/stub.tm.yaml

# import a public NOMAD archive entry (emits spec: tm-spec/0.3 with a
# G09_geometry_origin gate: dft_relaxed for relaxations, dft_static for
# single points, mlip_relaxed for ML methods; prodromos-ready out of the box)
tm-spec import-nomad <entry_id> --out imported.tm.yaml

# import structures from any OPTIMADE provider (Materials Project, NOMAD,
# OQMD, Alexandria). Structure-level only, so geometry_origin is honestly
# "unknown" and calculation is a minimal DFT stub.
tm-spec import-optimade --elements Fe S --provider mp --out fes.tm.yaml
tm-spec import-optimade --reduced-formula FeS2 --json

# import the COMPUTED magnetic ground state from Materials Project (the magnetic
# DEPTH that OPTIMADE/NOMAD lack): ordering + per-site magmoms -> magnetic block.
# Needs a (free) MP_API_KEY env var. AFM subtype unspecified by MP -> AFM-G + warning.
tm-spec import-mp --formula FeS2 --space-group 205 --out pyrite.tm.yaml
tm-spec import-mp --material-id mp-226 --json

# import an EXPERIMENTAL magnetic structure from MAGNDATA (Bilbao) by entry code.
# Experimental ordering anchor for MP's computed magnetism (which can be wrong, e.g.
# troilite/chalcopyrite). FM/AFM derived from the magCIF's own magnetic symmetry ops.
tm-spec import-magndata --code 0.1 --out lamno3.tm.yaml
tm-spec import-magndata --code 0.1 --mcif local_entry.mcif --json   # parse a local .mcif

# merge two TM-Spec docs locally (fill-only, same-material guarded): base
# depth (e.g. NOMAD method/results) is kept, overlay fills the holes (e.g.
# OPTIMADE formula variants + lattice_vectors).
tm-spec merge nomad.tm.yaml optimade.tm.yaml --out merged.tm.yaml

# one-shot: import OPTIMADE and merge each hit into a local NOMAD base
tm-spec import-optimade --reduced-formula FeS2 --merge nomad.tm.yaml --out merged.tm.yaml

# diff a hand-crafted pilot against its paired script
tm-spec lint examples/pyr_smoke.tm.yaml

# auto-fill sanity gates from run artefacts
tm-spec sanity-fill examples/pyr_smoke.tm.yaml \
    --json results/neb_canonical_pyr.json \
    --xyz  results/relaxed_pristine.xyz \
    --out  filled.tm.yaml

# build a NOMAD upload bundle
tm-spec export-nomad examples/*.tm.yaml --out nomad_upload.zip

Without install: python -m tm_spec.cli ... after pip install -e ..


A 30-second TM-Spec example

spec: tm-spec/0.2
kind: NEBCalculation
id: tm.pyr.vs.hint.smoke.2026-04-29
schema_url: https://exopoiesis.github.io/tm-spec/0.2.json

structure:
  formula: Fe32S63H1
  prototype: AB2_cP12_205_a_c
  space_group: { number: 205, symbol: Pa-3 }
  cell: { a: 5.418, c: 5.418 }
  supercell: [2, 2, 2]
  pbc: [true, true, true]

calculation:
  method: DFT
  level:
    xc: PBE+D3(BJ)
    basis: { kind: plane_waves, cutoff_Ry: 60, rho_cutoff_Ry: 240 }
    smearing: { kind: gaussian, width_Ry: 0.005 }
    spin: none
  k_points: { mesh: [2, 2, 2] }
  code: { name: QuantumESPRESSO, version: 7.3.1 }

workflow:
  kind: NEB
  stage: smoke
  endpoints:
    A: { ref: artifacts/endA.extxyz, E_eV: -128055.5404, fmax: 0.027 }
    B: { ref: artifacts/endB.extxyz, E_eV: -128055.5402, fmax: 0.026 }
  n_images: 7
  optimizer: BFGS
  prewrap: idpp

results:
  status: PASS
  paper_quotable: false

sanity:
  - { id: G01_FeS_bond, rule: "min(Fe-S) > 2.00 A", observed: 2.27, pass: true }
  - { id: G04_fmax_endpoints, rule: "fmax(A,B) <= 0.05", observed: [0.027, 0.026], pass: true }

provenance:
  date: 2026-04-29
  author: igor@exopoiesis.space
  parents: ["tm.pyr.vs.hint.pristine@2026-04-29"]
  compute: { host: cloud-gpu, gpu: A100, cost_usd: 4.0, walltime_h: 6.0 }
  hash:
    inputs: sha256:placeholder_to_be_computed
    outputs: sha256:placeholder_to_be_computed

Supported kinds

kind Required sections Pilot
NEBCalculation workflow, results examples/pyr_smoke.tm.yaml, mack_vfe_neb.tm.yaml
USCalculation cv_definition, sampling, pmf_analysis, results examples/w2_us_pmf.tm.yaml
MetaDynCalculation cv_definition, metadyn_protocol, fes_analysis, results examples/w2_metad.tm.yaml
MDCalculation md_protocol, results examples/w1_grotthuss_aimd.tm.yaml
MLIPBenchmark benchmark_setup, metrics, results examples/w2_mlip_benchmark.tm.yaml
SinglePointCalculation results examples/nomad_pyrite_singlepoint.tm.yaml
RelaxCalculation relax_protocol, results examples/nomad_relax_example.tm.yaml
Structure, Defects, Magnetic, Environment, Reaction, SanityReport, Provenance compositional fragments for JSONL streams n/a

Optional sections (any kind):

section What it records Pilot
preflight predictive pre-flight assessment from an external engine, BEFORE the run (counterpart of post-hoc sanity); shares the gate vocabulary in docs/gate-registry.md examples/preflight_example.tm.yaml

Repository layout

tm-spec/
├── README.md, SPECIFICATION.md, CHANGELOG.md, CITATION.cff
├── LICENSE              # MIT (code)
├── LICENSES/            # MIT + CC-BY-4.0 texts
├── pyproject.toml
├── schemas/
│   ├── 0.1.json
│   ├── 0.2.json
│   └── 0.3.json
├── docs/
│   ├── specification/v0.1.md
│   ├── specification/v0.2.md
│   ├── specification/v0.3.md
│   ├── gate-registry.md
│   ├── design-decisions.md
│   ├── standards-alignment.md
│   ├── landscape.md
│   └── lit-review.md
├── examples/
├── src/tm_spec/
└── tests/

Tests

pip install -e ".[dev]"
pytest -q
ruff check src tests
tm-spec validate --all examples --strict

CI runs on Python 3.10, 3.11, 3.12, and 3.13.


Versioning

tm-spec/<major>.<minor> is part of each YAML document (spec: field). Schema files are versioned in schemas/<major>.<minor>.json. Breaking schema changes bump <major>; additive fields bump <minor>. Code follows SemVer independently (pyproject.toml).

Spec Package Status Highlights
0.1 0.1.0 DRAFT Initial 11 kinds, 6 pilots, validator/extract/lint/sanity-fill/NOMAD export
0.2 0.2.0 DRAFT previous NOMAD importer, SinglePointCalculation, RelaxCalculation, import provenance
0.3 0.3.0 DRAFT current Optional preflight block (predictive) + endpoint.geometry_origin + shared gate registry

Citation

If TM-Spec helped your paper SI, cite:

Morozov, I. (2026). TM-Spec: a declarative YAML metalanguage for reproducible atomistic calculations. Version 0.3. https://github.com/exopoiesis/tm-spec


License

This project is dual-licensed using the REUSE/SPDX convention:

  • Code (src/, tests/, build files) — MIT.
  • Specification text & schema (docs/, schemas/, SPECIFICATION.md) — CC-BY-4.0.

Root LICENSE is a copy of MIT so GitHub detects the primary code license.


Contact

Author: Igor Morozov, igor@exopoiesis.space
Issues: https://github.com/exopoiesis/tm-spec/issues

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

tm_spec-0.3.0.tar.gz (99.4 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

tm_spec-0.3.0-py3-none-any.whl (92.4 kB view details)

Uploaded Python 3

File details

Details for the file tm_spec-0.3.0.tar.gz.

File metadata

  • Download URL: tm_spec-0.3.0.tar.gz
  • Upload date:
  • Size: 99.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for tm_spec-0.3.0.tar.gz
Algorithm Hash digest
SHA256 ccbd9e4c7bcab3ce945a969d43338270ee17f9ffa010b1ca648af6e7d0892ff8
MD5 dfd15327373fe743b1dc98c6b4eb69f2
BLAKE2b-256 4e9ecf79ede314f3851913c9329aca1fbfe746a34eab041b3334e0e3d72dc6a3

See more details on using hashes here.

Provenance

The following attestation bundles were made for tm_spec-0.3.0.tar.gz:

Publisher: release.yml on exopoiesis/tm-spec

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file tm_spec-0.3.0-py3-none-any.whl.

File metadata

  • Download URL: tm_spec-0.3.0-py3-none-any.whl
  • Upload date:
  • Size: 92.4 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for tm_spec-0.3.0-py3-none-any.whl
Algorithm Hash digest
SHA256 fcc830a556e05c414345ead6702c84492d7a536e84748ca65687e66c3f2c0dfb
MD5 c16511e47d5d55f116965e1ac2723c69
BLAKE2b-256 bd4e5b12ee8d424bc6cb74ceb8412b8ac7dbfaeb88f9e9125483d1a1e66f4a5a

See more details on using hashes here.

Provenance

The following attestation bundles were made for tm_spec-0.3.0-py3-none-any.whl:

Publisher: release.yml on exopoiesis/tm-spec

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page