README.md

TNFR examples

Examples demonstrate declared constructions, APIs and conditional results. Their numbers and historical filenames are discovery aids, not a ranking of scientific validity. The theory index owns claim status; the execution plan owns the active research queue. Running an example does not reopen a parked branch.

The curvature safety illustration compares supplied WS-graph field observations before and after operator events. Its fit and safety labels are configured advisories, not a stability theorem. It moved from benchmarks/; the single-use topology helper was removed while preserving the declared preparation.

Start with the intended task

Directory	Use and authority	Interpretation
01_foundations	SDK and grammar scope	Supplied preparations, words and state ensembles
02_physics_regimes	Diffusion certificates, diagnostics and auxiliary models	Read each model's hypotheses; a diagnostic decrease is not general stability
03_riemann_zeta	Finite instruments in the Riemann notebook	Parked comparisons; disclose supplied zeros/primes
04_riemann_L_twisted	Character/L-function instruments in the same notebook	Supplied arithmetic data and finite comparisons, not generalized RH
05_type_hygiene	Operator and state-type scope	Finite delay projection, actual storage and event/window observations
07_number_theory	Arithmetic definitions, residues and prime structure	Disclose factorization, sieves and other construction inputs
08_emergent_geometry	Scale/geometry bridge, spectra and auxiliary models	Separate prescribed geometry, observed structure and autonomous generation
10_applications	Measurement protocol, adapters and backend provenance	Data admission and reserved prediction remain separate obligations

Use rg --files examples -g "*.py" for the current file inventory. The grammar automaton examples reuse _flat_grammar_model.py; arithmetic and physical examples reuse their package owners. A finite fixture may recur as a controlled comparison without constituting a second implementation of its governing law.

Install the repository before running a selected entry point:

python -m pip install -e .
python examples/01_foundations/01_hello_world.py
python examples/01_foundations/04_operator_sequences.py
python examples/01_foundations/10_simplified_sdk_showcase.py

Example 01 is the minimal SDK entry point. Example 04 separates flat grammar admission from a deliberately illustrative pressure proxy. Example 10 surveys the topology builders and configured SDK operations. Example 07 is a larger prepared-state ensemble, not part of this introductory command list.

For a retained declaration and finite execution report, use reproducible_study.py and the shared CLI/SDK guide. The same declaration can be run through either interface; its export is not a complete resumable checkpoint.

Optional dependencies vary by script. Plotting examples require viz-basic; examples 91 and 92 additionally require scikit-learn, and 92 downloads and caches UCI data. The Torch demonstration uses compute-torch. Read module docstrings and available --help before an explicit run; the directory is not an instruction to execute every file. Importing a demonstration does not start its command-line workflow. Some older examples still configure imports or plotting defaults at module scope, so this is not a promise of side-effect-free imports. Package ownership belongs to Architecture, and verification requirements to Testing.

Retained controls and shared owners

The type-hygiene directory retains these scoped entry points:

Entry	What it observes
77	Finite fixed-delay Fourier projection and window sensitivity
79	Actual scalar-chart storage membership and descriptive temporal entropy
82	Selected finite REMESH event/window comparisons

Entropy thresholds and scalar-only fixtures do not prove that a richer state type is necessary or impossible. The retired type-necessity and synthetic closure campaigns are replaced by the actual domain and execution contracts: signed-form admission, circular U3 admission, field availability and grammar observations.

For supplied topology construction use 10; for declared diffusion use 99. Prescribed phase/form response belongs to 179. Removed chemistry and Millennium demonstrations supplied target laws or encodings without deriving them from nodal dynamics. Their removal does not affect the scoped diffusion, graph-wave, winding and arithmetic owners. Retirement reasons and source recovery belong to the recovery index; interface changes belong to the API migrations.

Diffusion and runtime evidence

These examples reuse the diffusion stability owner. The table distinguishes an exact declared model from finite engine evidence. Detailed bounds and assumptions remain in that owner instead of being duplicated as a second theorem ledger here.

Entries in 02_physics_regimes/	What they demonstrate	Boundary
160, 161, 162	Restricted S16 and affine-reset controls	Declared pure-EPI/hybrid models
163, 164, 165	Local operator and event-time binding	Captured runtime maps, not arbitrary operators
166, 167, 168	Modal reference, Euler refinement and finite binary64 defects	Conditional exact-real refinement differs from runtime convergence
169, 170	Causal cycle receipts and finite block margins	One invocation/block does not certify future stability
171, 172, 173	Conditional history envelopes and represented-number defects	Required gain/defect hypotheses are separate from their finite measurements
174, 175, 176, 177	Restricted P2 half-Reception and alpha-one REMESH composition	Numeric kernels, one stage, finite sequence and revalidated policy have distinct scopes
178	Restricted antisymmetric binary64 REMESH class	REMESH invariance alone is not complete-runtime invariance

Geometry and phase/form response

Example 180 runs four small joint steps of the explicitly selected capacity-separable relational model, including a zero-capacity node. It calls the shared SDK/engine owner and prints actual storage increments and Euler defects. It also reports regional work/cut accounting and unweighted phase-response covariance with their exact arithmetic evidence, including at zero capacity. The supplied preparation, storage scale and capacity premise remain explicit; this is no formation or physical-validation claim. A separate read-only block applies all three capture certificates to supplied asymmetric two-ring snapshots. It shows why an unavailable exact-reflection certificate can coexist with admitted full-state local and sector certificates, and why positive heterogeneous capacity and a nonunit storage scale require the broader sector theorem. This comparison runs no additional trajectory and revises no frozen result. See the execution guide.

python examples/08_emergent_geometry/180_relational_exchange.py

The scale bridge owns effective geometry and observation closure. The tetrad is a required diagnostic interface, not a proven complete state; reflection equivalence and loss of hidden state are separate questions. Auxiliary symplectic/U(2) demonstrations do not establish that engine operators are Hamiltonian or generate particles.

The unified-field showcase compares seeded supplied states on a path, cycle and barbell using the shared snapshot readout. It executes no operators or evolution. Its optional plot reports descriptive field statistics and explicitly leaves temporal conservation unavailable; the former handwritten operator substitutes and physical-domain validation claims have been removed.

Example 179 checks a prescribed rotating phase contrast and its derived EPI response on the six-node prism. It evaluates detached analytic snapshots rather than a native trajectory. The phase/form owner states the exact assumptions, moving mean and same-input contraction scope. The imposed phase clock is not a derived autonomous maintenance mechanism.

python examples/08_emergent_geometry/179_phase_form_driven_response.py --output-dir output/phase_form_driven_response

Optional plots require the viz-basic extra. Retained outputs are the figure, JSON and CSV. Their recorded residuals and refinements are finite evidence, not autonomous formation or a physical identification.

The offline interface example compares a planted spatial target and supplied sinusoidal blocks using shared observation adapters. It executes no coupling law or nodal trajectory. The signal adapter currently omits coherence-length fit/fallback provenance; the example exposes that limitation and treats operator suggestions as telemetry.