Merge relational coefficient and research consolidation

Author: TNFR AI Agent

ARCHITECTURE.md

@@ -53,6 +53,8 @@ graph invalidation and persistence retain their separate responsibilities.
 | Regional orientation relative to a supplied rate source | [`source_relative_form.py`](src/tnfr/physics/source_relative_form.py); reuses the regional form observer |
 | Conditional relational phase/form execution | [`dynamics/relational.py`](src/tnfr/dynamics/relational.py); one admitted field and atomic Euler step, reused by the SDK |
 | Prepared relational pattern observations | [`physics/relational_observations.py`](src/tnfr/physics/relational_observations.py); supplied regions/reference lifts, shared winding and regional support accounting |
+| Conditional coefficient uncertainty | [`physics/relational_observations.py`](src/tnfr/physics/relational_observations.py); exact three-sample stencils delegate to shared outward jet bounds, with declared preparation/clock/error premises |
+| Retained P2 acquisition audit | [`research/relational_acquisition.py`](src/tnfr/research/relational_acquisition.py); read-only protocol/archive and all-step error reconstruction; does not import the producer or evolve a graph |
 | Protected relational basins | [`physics/relational_capture.py`](src/tnfr/physics/relational_capture.py); reflected, full-state local and acute-sector theorems sharing exact phase/energy enclosures for conditional ideal-law limits |
 | Validated relational transit | [`physics/relational_transit.py`](src/tnfr/physics/relational_transit.py); exact reflected ODE enclosure using shared rational intervals, Taylor derivatives and signed-diagonal comparison; read-only proof computation, not live engine evolution |
 | Coherence and equilibrium kernel | [`common.py`](src/tnfr/metrics/common.py) |
@@ -220,6 +222,25 @@ thin adapter; the SDK's `relational_report_to_dict` supplies the exact rational
 JSON projection for the existing writer. Reports are observations, not live
 checkpoints or authenticated execution histories.
 
+The same observation owner exposes graph-independent coefficient-jet and
+three-sample bounds. Exact rational stencil/error propagation delegates to the
+shared outward interval kernel and mathematical-pi enclosure, rather than a
+second coefficient estimator. These are module-level APIs, not `Network`
+methods: the caller supplies preparation, affine timing, baseline/gain and
+independent uncertainty. The existing exact report exporter supports both
+immutable report types. [Usage](docs/guides/REGIONAL_AND_RELATIONAL.md#bound-a-prepared-coefficient-response)
+and [admission](docs/contracts/RELATIONAL_DYNAMICS.md#relational-coefficient-samples)
+keep those external premises separate from arithmetic availability.
+
+`research/relational_acquisition.py` reconstructs the retained known-source P2
+record without invoking `dynamics/relational.py` or importing a benchmark.
+It independently checks the declared model/window, reads saved states and
+rates, verifies protocol/source-archive bytes and recomputes error budgets
+through the shared observation and interval owners. Internal consistency,
+the original precision verdict and missing evidence have separate outputs.
+It neither authenticates execution nor requires current source to match a
+historical archive; the frozen producer and original records remain unchanged.
+
 The same observer's `Network.relational_attachment` route compares two
 separately admitted components with their hypothetical joined field. Port
 messages, exact represented rate changes and shared support-reset accounting

README.md

@@ -139,6 +139,14 @@ The results connect several parts of this picture:
    proved recovery region. The result also admits sufficiently small changes
    to the preparation. It concerns one supplied change followed by the declared
    dynamics, not an automatically chosen sequence of changes.
+7. **A response can identify a coefficient without selecting a law.** For the
+   stated relational family, a prepared form mode and initial phase consensus
+   let value, rate and acceleration identify one coefficient combination.
+   Shared tools propagate uncertainty and abstain when it is unresolved. A
+   known-source two-node computational acquisition now has a whole-window
+   nonlinear and numerical error budget, rather than relying on a small solver
+   residual. It verifies the estimation procedure inside that model; different
+   nonlinear laws can still share the same initial response.
 
 Negative controls are useful parts of these results. For example, two
 preparations with the same model-defined initial energy can reach different
@@ -154,25 +162,38 @@ complete set of rules, check that the rules work together, and predict something
 not used to choose those rules. Reusable results are integrated into shared tools
 and the engine where their scope supports it, with explicit conditions for use.
 
-**The immediate question is what can select a change of connections and its
-timing.** We now have a conditional example in which the change respects both
-the storage budget and subsequent pattern recovery. The next study examines
-which choices the full nodal state, symmetries and clock actually constrain,
-and which would require an additional rule. It will not infer that a permitted
-event must happen or use node labels to break an unexplained tie.
-
-The budget alone does not select an event. Keeping a loop at the instant of
-a change also does not, by itself, guarantee later recovery. The research
-separates these obligations so that an assumed event schedule cannot be
-mistaken for a mechanism generated by the model.
-
-This step matters to the larger objective: explaining how coherent structures
-can compose and interact through their own dynamics, with fewer unexplained
-choices supplied from outside. The
-[execution plan](https://github.com/fermga/TNFR-Python-Engine/blob/main/theory/research/FIVE_STAGE_EXECUTION_PLAN.md#current-g3-gate)
-owns that next task; the
+**The current bounded question is how to distinguish constitutive laws after
+estimating a coefficient.** Two laws can obey the same storage balance and
+produce the same initial response around phase consensus, yet respond
+differently away from it. Calibration determines a setting; a separate,
+uncertainty-aware prediction must test which candidate response is compatible.
+Estimating the setting is not evidence that the law is true.
+
+The [coefficient study](theory/nodal/RELATIONAL_EXCHANGE_ADMISSION.md#coefficient-synergy-audit)
+connects coefficient freedom, hidden-phase memory and observable response.
+Its [prepared identification and bounded acquisition](theory/nodal/RELATIONAL_EXCHANGE_ADMISSION.md#prepared-coefficient-identification)
+provide shared engine observations and a verified known-source software control.
+The next specification must carry an admitted calibration interval into both
+named candidates without refitting, and establish a discriminating prediction
+or a precise obstruction before another trajectory. This remains an internal
+model-admission question, not a new physical-data campaign.
+
+The [effective-link and collective-memory results](theory/nodal/RELATIONAL_PATTERN_COMPOSITION.md#effective-link-admission)
+remain reusable: supplied nodal environments can transmit deformation and
+restore joint geometry, and reductions must retain their hidden information.
+The [connection-foundation results](theory/nodal/RELATION_FOUNDATIONS.md)
+separate primitive support from transport weight and effective interaction.
+They admit conditional contact mechanisms but do not select a birth law.
+Formation of the initial substrate, autonomous law selection and physical
+identification remain open; this does not make the valid conditional results
+obsolete or prove that those open goals are impossible.
+
+For the current repository checkpoint, deliverables and stopping rule, use the
+[execution plan](theory/research/FIVE_STAGE_EXECUTION_PLAN.md#current-g3-gate)
+as the sole task queue. Its [resumption map](theory/research/FIVE_STAGE_EXECUTION_PLAN.md#current-checkpoint)
+locates reusable results and evidence; the
 [research portfolio](https://github.com/fermga/TNFR-Python-Engine/blob/main/TNFR_lineas_de_investigacion.txt)
-explains the roles of the supporting studies.
+classifies the supporting branches without scheduling additional campaigns.
 
 ### How this could connect to physical reality
 
@@ -267,6 +288,7 @@ seeds; direct `TNFR.create(..., seed=...)` sets the topology seed.
 | Compare an attachment | `relational_attachment` | Fresh separate/joined fields for a hypothetical connection; does not change live support |
 | Compare a bridge relocation | `relational_relocation` | Supplied atomic bridge exchange preserving internal support; passive budget and recovery are separate obligations |
 | Apply a capture or transit theorem | `Network.relational_*capture` | Read-only sufficient certificates with support-specific hypotheses |
+| Bound a prepared coefficient response | `bound_relational_coefficient_from_jet`, `bound_relational_coefficient_from_samples` | Module-level observers with declared uncertainty; no graph evolution or preparation authentication |
 
 The [regional and relational guide](https://github.com/fermga/TNFR-Python-Engine/blob/main/docs/guides/REGIONAL_AND_RELATIONAL.md)
 provides preparations and examples for joint dynamics. The
@@ -300,6 +322,7 @@ belong to the [structural field guide](https://github.com/fermga/TNFR-Python-Eng
 | What constrains a change of connections? | [Event storage, passivity and selection limits](https://github.com/fermga/TNFR-Python-Engine/blob/main/theory/nodal/RELATIONAL_PATTERN_COMPOSITION.md#support-event-premise-admission) |
 | Can a pattern survive a relocated connection? | [Passive relocation and recovery](https://github.com/fermga/TNFR-Python-Engine/blob/main/theory/nodal/RELATIONAL_PATTERN_COMPOSITION.md#identity-preserving-bridge-relocation) |
 | How does hidden state produce memory? | [Derived pattern memory](https://github.com/fermga/TNFR-Python-Engine/blob/main/theory/nodal/RELATIONAL_PATTERN_MEMORY.md) |
+| What can identify a coefficient, and what does not select its law? | [Coefficient, memory and preparation admission](theory/nodal/RELATIONAL_EXCHANGE_ADMISSION.md#coefficient-synergy-audit) |
 | Which terms are assumptions, results or observations? | [Classified glossary](https://github.com/fermga/TNFR-Python-Engine/blob/main/theory/GLOSSARY.md) |
 | How will physical claims be tested? | [Measurement and reserved evaluation](https://github.com/fermga/TNFR-Python-Engine/blob/main/theory/research/FIVE_STAGE_EXECUTION_PLAN.md#supporting-measurement-bridge) |
 
@@ -317,9 +340,11 @@ does not make them interchangeable with the main generative model.
 | `src/tnfr/physics/`, `mathematics/`, `metrics/` | Theorem tools, numerical algebra, observations and diagnostics |
 | `src/tnfr/sdk/`, `cli/` | Public networks, study recipes, reports and command adapters |
 | `src/tnfr/config/`, `constants/`, `utils/` | Shared configuration, numerical settings, validation helpers and I/O |
+| `src/tnfr/research/` | Reusable admission, provenance and read-only record audits; not another evolution runtime |
 | `theory/` | Mathematical definitions, derivations, research strategy and the execution plan |
 | `docs/` | Usage guides, execution contracts and documentation navigation |
 | `tests/`, `examples/`, `benchmarks/` | Contract checks, executable illustrations and scoped measurement/research instruments |
+| `docs/assets/`, `artifacts/` | Preserved published evidence and local protocols/source archives/responses; missing local evidence remains unavailable |
 | [applications/](https://github.com/fermga/TNFR-Python-Engine/blob/main/applications/README.md) | Optional arithmetic applications with explicit verification boundaries |
 
 [Architecture](https://github.com/fermga/TNFR-Python-Engine/blob/main/ARCHITECTURE.md)

TESTING.md

@@ -63,7 +63,10 @@ second inventory of research results or individual test cases.
 The routine gate includes regional/source-relative form observations, generic
 linear observation admission, relational field/step execution, pattern,
 attachment and relocation observations, work/cut/phase-response accounting,
-protected capture and the listed current production boundaries. The attachment
+protected capture and the listed current production boundaries. Coefficient
+jet/sample observers and read-only acquisition-audit admission are
+also covered by the routine top-level modules. The retained full-window record
+is an optional research audit, not an implicit producer invocation. The attachment
 observer evaluates two components and a hypothetical joined field; relocation
 compares the admitted fields before and after a supplied bridge exchange.
 Neither observer changes live support or certifies future pattern recovery.

TNFR_lineas_de_investigacion.txt

@@ -1,6 +1,6 @@
 TNFR - RESEARCH PORTFOLIO
 ========================
-Updated: 2026-09-27.
+Updated: 2026-10-02.
 
 1. PURPOSE AND DOCUMENT RESPONSIBILITIES
 ========================================
@@ -43,9 +43,26 @@ identity alone does not determine pressure, phase, capacity, support or events.
 Source-relative observations and fixed-law closure criteria now have shared
 engine/SDK owners. They retain their supplied-source and domain assumptions.
 
+Connection foundations now distinguish declared support, normalized transport
+weight, observation distance and causal dependence. A zero-weight edge is not
+necessarily an absent relation, and first-neighbor normalization differs from
+contact between already active regions. The relation state and zero-boundary
+admissions retain distinct autonomous passive contact completions, without
+selecting relation-only storage or kinetics. Fixed-support nodal evolution
+cannot determine them. These are reusable foundational constraints, not an
+active parallel creation-law search.
+
+The coefficient audit likewise separates conditional model consistency from
+constitutive selection. Balance, independent capacity composition, clock
+covariance and synchronized replicas permit different chi values. Prepared
+response and memory can identify chi within the admitted two-channel family;
+they do not derive a universal coefficient or reduce all TNFR freedom to it.
+The execution plan alone sets the current calibration/discrimination gate.
+
 Primary owners:
   theory/FUNDAMENTAL_THEORY.md
   theory/NODAL_PARAMETER_FOUNDATIONS.md
+  theory/nodal/RELATION_FOUNDATIONS.md
   theory/nodal/PRESSURE_CONSTITUTIVE_SCOPE.md
   theory/nodal/JOINT_PARAMETER_RESPONSE.md
   theory/nodal/DERIVED_FORM_PHASE.md
@@ -63,6 +80,13 @@ It has shared opt-in engine/SDK execution, signed-work and regional-response
 observations, and separate sufficient capture/transit certificates. The
 default operator runtime remains a different execution path.
 
+Coefficient-jet and three-sample uncertainty observers now share an engine
+owner and exact report export, with explicit abstention. One known-source P2
+temporal acquisition has a prior-derived nonlinear derivative bound and a
+retained numerical error chain. This verifies the procedure within the model,
+not blind calibration, a physical measurement or preference for one same-jet
+nonlinear completion. Its records remain immutable and independently auditable.
+
 On supplied two-ring support, a validated continuous transit joins formation
 to a protected maintained phase geometry. Controls with uniform or reversed
 initial form reach consensus under the same law. Initial energy and loss
@@ -94,24 +118,35 @@ improved both tangent and direct-cubic controls on matched Euler grids.
 That result neither certifies an exact-ODE error at the evaluated amplitude
 nor supplies an autonomous compressed node or a fitted memory law.
 
-The component-to-joined interface is now admitted for one supplied acute unit
-bridge, retaining full internal state. Degree, internal form gradient and the
-complex neighbor resultant suffice for each new port row. Equal endpoint
-form/phase and zero event storage do not imply unchanged local rates, excluding
-the endpoint-difference-only control. The shared read-only engine/SDK report
-retains actual represented changes; it does not create a connection.
-Support-event accounting and nonselection are now explicit. With unchanged
-nodal state, pure addition costs nonnegative joint storage. An additional
-zero-supply passivity premise admits only equal form/phase endpoints and does
-not select occurrence; a continuous addition intensity on an open full-state
-domain must vanish. A supplied atomic edge exchange can lower storage on an
-open set. An explicit two-C5 bridge relocation now retains both internal
-cycles and a shared sufficient continuous recovery basin on a strict open
-domain. Its reverse can preserve recovery while failing the passive budget;
-the zero-perturbation control is a neutral exchange of equilibria.
-The shared observer/SDK compares the event without modifying live support.
-The remaining dependency is constitutive event choice and clock: a favorable
-budget and recoverable identity do not force a connection change to occur.
+Retained mechanisms in this branch are classified by what they preserve:
+- Attachment and relocation retain port neighborhoods, full state and event
+  work. A passive change and subsequent recovery have separate hypotheses;
+  neither selects the event or its time.
+- Joint resets retain simultaneous nodal/support changes. Actual UM and
+  RA-then-UM witnesses are conditional budget results, not autonomous birth.
+  The all-operator map owns configured actions and activation boundaries.
+- Mediation retains joint form/phase memory and hidden initial information.
+  A frozen intermediary can restore geometry without transmitting changes;
+  a causal-null control is therefore necessary for an effective relation.
+- Effective links combine mutual influence with local restoring geometry on
+  supplied fine support. Return paths and damped modes add scoped geometric
+  information; a visible or permanent pulse is not required.
+- Fast and series reductions retain selected phase lifts, original port
+  normalization, inherited storage and the initial transient. Stationary
+  composition and finite-capacity evolution have distinct error obligations.
+- Three-port reduction retains genuinely collective phase information;
+  independent pair interactions cannot generally replace the whole law.
+- Pressure-state counterexamples retain the environment's hidden coordinates.
+  Zero pressure is not absence of structure, and a physical vacuum remains
+  an unestablished interpretation.
+
+These results are linked from their owners below and the theory-to-execution
+map, rather than copied into a delivery chronology. Primitive support origin,
+unrestricted nonlinear composition and changes to another winding identity
+remain separate open/deferred questions. Physical analogies are prompts, not
+validated constituents or additional campaigns. The current coefficient and
+same-jet-law distinction uses these observation lessons without reopening
+their completed producers.
 Precise obligations and acceptance criteria belong only to:
   theory/research/FIVE_STAGE_EXECUTION_PLAN.md#current-g3-gate
 
@@ -220,7 +255,9 @@ Preserve original model, inputs, source, precision, clock and frozen outcomes.
 A later theorem or code correction does not rewrite an earlier verdict.
 
 Autonomous support/capacity selection, unrestricted nonlinear composition and
-physical identification remain open. A restricted positive result need not
+physical identification remain open. The completed support-selection audit
+does not reopen an event-law search without an independently justified premise.
+A restricted positive result need not
 explain creation from nothing, prove infinite lifetime for every pattern or
 select every operator universally. Resume at the plan's current gate, using
 the substantive owners above and the archive as evidence, not as instructions.