TNFR 0.0.3.6 - Nodal research integration and documentation consolidation

TNFR 0.0.3.6 — Nodal research integration and documentation consolidation

This release integrates the nodal research and implementation work since
0.0.3.5 and brings the documentation into agreement with the implemented
contracts. It makes the distinction between a derived identity, a conditional
theorem, a finite runtime observation and an open hypothesis explicit.

Nodal dynamics and effective descriptions

• Shared owners now connect signed scalar EPI, phase/form observations,
  support transport, held forcing, quotients and derived memory. Scalar-only
  paths reject nonuniform or complex EPI representations instead of silently
  substituting a magnitude for a signed form coordinate.
• Exact affine realization supports declared joint outputs and offsets. The
  geometry adapter keeps model predictions, captured pressure and numerical
  residuals separate. These are conditional descriptions of specified laws,
  not a mechanism that selects those laws autonomously.
• The retained five-node path has a complete reflection-invariant form
  description and a constrained exact decoder. Removing the reflection label
  does not remove its hidden continuous degrees of freedom. Field covariance,
  omitted-form effects and correlation-estimator limitations are recorded.
• Diffusion, finite REMESH/Reception execution, THOL birth and regional exchange
  results retain their individual support, capacity, history, arithmetic and
  provenance conditions. Finite execution evidence is not promoted to a
  guarantee for arbitrary future operator sequences.

One documentation structure

The theory index
identifies the current owner of each result, and the
execution plan
is the sole research queue. Superseded proposals and development histories are
preserved as archives instead of competing with current instructions.

The release also consolidates contributor, testing, workflow and architecture
guides; generates the operator-contract table from its registry; validates
documentation references and build inputs; and uses one MkDocs toolchain.
The educational manuscript and its site navigation have been removed.
The root README now provides one installation path, an explicit checked operator
example and navigation to the shared CLI/SDK guide. Agent guidance identifies
the shared execution and diagnostic owners, recipe/report limits and reusable
test controls. Maintained reference headers link to package metadata instead
of repeating obsolete software versions.
Dead workflow/profiler entry points and obsolete benchmark routes were
retired with their original material and retirement records retained.

CI retains per-Python JUnit reports. CLI lineage checks use self-contained
synthetic bytes, and spectral checks compare against an analytic reference
without requiring bitwise equality from equivalent floating-point solvers.
Signature readouts isolate spectral-cache writes from the caller's graph,
including calls without a synthetic probe.

The test/example audit retires toy assertions, copied verifier criteria,
duplicate runners and obsolete curvature/timing campaigns. Introductory
examples use the shared SDK. CLI wiring is checked independently of expensive
numerical producers; file-grouped parallel tests reuse module fixtures.
Temporal reports reuse the evaluation's window count, and aggregate test extras
reference their existing groups without changing dependency requirements.
The retirement map records replacement controls and original source hashes.

The standalone primality package's advanced CLI now keeps progress messages
off JSON standard output, emits a single JSON document for its reporting
routes, and accepts the existing standard and advanced validation-result
schemas. Its help describes finite arithmetic comparisons without presenting
them as validation of the physical paradigm.

Shared CLI and SDK studies

StudySpec, run_study and diagnose_network provide one path for declaring
small network studies and reporting their actual observations. The CLI delegates
to these same owners through tnfr network; tnfr operators and tnfr sequences
read the existing registries without executing a study. python -m tnfr provides
the same commands without depending on a shell entry-point lookup.

JSON recipes and reports are separate artifacts. Reports preserve initial state,
software versions, stored-pressure scope, unavailable observations and coherence
length provenance. Cycles count complete requested operator words, not physical
seconds; inherited configuration and permitted runtime substitutions remain
explicit. A recipe is not a complete checkpoint or a cross-version replay promise.

The CLI and SDK guide
and executable example cover construction, execution, export and replay. Logs
use stderr without replacing the caller's logging configuration. JSON stdout
works with Windows legacy encodings; each output file uses the shared atomic
writer. Unavailable old profiling commands have been removed.

Scientific scope and the next step

The nodal relation dEPI/dt = nu_f * DeltaNFR specifies form rate once its
inputs are supplied. It does not independently determine pressure, phase,
capacity, support or autonomous operator selection. The structural tetrad is
a diagnostic interface, not a proven complete state basis. Auxiliary wave,
Hamiltonian and arithmetic models have their own assumptions.

The next active G3 task is a controlled approximation: bound the pressure and
specified field errors caused by omitting decaying hidden form on the existing
five-node path. C6 global stability remains unresolved and parked; it is not a
second active queue. An independently admitted terrestrial measurement model
and reserved empirical prediction also remain open.

This release does not establish physical particle emergence, the emergence of
observed reality, unrestricted network stability or a Millennium conjecture.
It provides more precise models, reusable mathematical results and executable
controls for evaluating those research questions.

Release validation

CI on the release commit
passed on Python 3.10, 3.11, 3.12 and 3.13. Each interpreter reported 13,934
passed tests, 798 skipped tests and zero failures or errors. Python 3.11 line
coverage was 77.88% (76,257 of 97,915 lines); branch coverage was not measured.
The default selection excludes marked slow research producers, and optional
compute backends are covered only where installed. The focused SDK suite passed
363 tests; these overlap the general suite and are not additional unique coverage.
Documentation, internal references and the dependency audit also passed on the
same commit. Advisory static analyses remain separate from enforced checks.

The reviewed wheel and source distribution contain 733 package files identical
to the release sources. A fresh wheel installation passed dependency checks and
exercised the README example, shared CLI/SDK reports, recipe export and replay,
CPU FFT, read-only signatures and empty spectral sectors. Both distributions
passed Twine checks.

The published documentation was checked against the release commit, including
the README, agent guidance, architecture, tetrad and CLI/SDK guide. Historical
research evidence remains unchanged. Attached validation receipts and SHA-256
checksums identify the reviewed artifacts and the scope of these checks.

Upgrade and references

SciPy is now declared as a core dependency, matching the SDK's existing spectral
path. Optional psutil memory telemetry can be unavailable without blocking
the CPU computation interfaces.

Empty disagreement sectors have explicit matrix-exponential and transient-gain
results across supported NumPy versions. Terrestrial timestamp ingestion accepts
UTC Z and variable-width fractional seconds consistently on Python 3.10,
without rounding or changing stored timestamps, while retaining unavailable and
invalid-time statuses. FFT phase-preservation checks compare circular separation,
including the zero/2-pi branch, with the existing numerical tolerance.

python -m pip install "tnfr==0.0.3.6"

Review the API contracts
when upgrading. Strict scalar admission, unavailable phase-curvature outcomes
and correlation-estimator provenance are meaningful parts of the interface.

Documentation ·
Source at this release ·
Changelog ·
Zenodo version family