fibernet 4.2.0__tar.gz → 4.4.0__tar.gz

fibernet-4.4.0/.gitattributes

@@ -0,0 +1,86 @@
+# The numerical cache identity hashes exact source bytes.
+FiberScope/fslab/*.py -text
+
+# Source-contract and native-core bindings require byte-identical checkouts.
+fibernet/sim/accelerated.py -text
+fibernet/sim/_energy_accounting.py -text
+fibernet/viz/__init__.py -text
+tests/fixtures/observer_prior71_accelerated.txt -text
+
+# Public regression baselines retain their source UTF-8 bytes.
+tests/fixtures/public_source_contract/**/*.txt -text
+
+# The retained CG acceptance record pins exact compiler bytes.
+fibernet/integrations/_lammps_cg.py -text
+
+# Exact API contract snapshots and their facade sources must keep Git bytes.
+fibernet/__init__.py -text
+fibernet/easy.py -text
+fibernet/api.py -text
+tests/fixtures/public_api_contract_v2/**/*.txt -text
+tests/fixtures/result_identity_contract_v1/compute_detailed.py.txt -text
+tests/fixtures/spring_coefficient_provenance_v1/accelerated_delta.json -text
+
+# Reviewed optional facade sources and exact source snapshots retain Git bytes.
+fibernet/sim/__init__.py -text
+tests/fixtures/optional_facades_v2/*.txt -text
+tests/fixtures/optional_facades_v2/sim_lammps_capabilities_delta.json -text
+tests/fixtures/optional_facades_v2/root_fiber_network_profile_delta.json -text
+adapters/mcp/tests/fixtures/provenance_v1_outputs.json -text
+
+# The reviewed explicit stretch-ratio API fixture retains exact source bytes.
+tests/fixtures/easy_stretch_ratio_v1/easy.py.txt -text
+
+# The reviewed random-rollout failure fixture retains exact source bytes.
+tests/fixtures/easy_rl_failure_v1/easy.py.txt -text
+
+# Atomic graph edits are bound to their exact reviewed source bytes.
+fibernet/core/structure_graph.py -text
+
+# The saved-temperature transfer owner is included in exact native-core bindings.
+fibernet/sim/saved_temperature.py -text
+
+# Thermoelastic API and owner are included in exact native-core bindings.
+fibernet/sim/thermoelastic.py -text
+fibernet/sim/_thermoelastic.py -text
+
+# Reviewed geometry-policy deltas retain exact UTF-8 bytes.
+tests/fixtures/taichi_geometry_policy_v1/*.json -text
+
+# Exact native-core bindings cover every package Python source.
+fibernet/**/*.py -text
+
+# Bundled saved-electrothermal containers retain their exact encoded bytes.
+tests/fixtures/saved_electrothermal_v1/*.gz.b64 -text
+
+# The deformation preparation golden retains its pinned UTF-8 bytes.
+tests/fixtures/deformed_transport_preparation_v1/planar_golden.json -text
+tests/fixtures/deformed_transport_preparation_v1/arithmetic_reference.py.txt -text
+
+# This retained scientific witness is verified by its exact original SHA-256.
+# Preserve repository bytes even when a checkout uses core.autocrlf=true.
+/tests/fixtures/dynamic_viscoelastic_generalized_summary.json -text
+tests/fixtures/optional_facades_v2/sim_archived_ge_delta.json -text
+tests/fixtures/easy_show_dimension_v1/easy_delta.json -text
+tests/fixtures/native_io_dispatch_v1/api_delta.json -text
+tests/fixtures/easy_batch_report_v1/easy_delta.json -text
+
+# Exact additive octet facade delta preserves its reviewed Git bytes.
+tests/fixtures/standard_octet_v1/root_facade_delta.json -text
+
+# Reviewed legacy summary facade delta.
+tests/fixtures/legacy_analysis_v1/api_delta.json -text
+
+# Exact reviewed LAMMPS persistence source delta.
+tests/fixtures/lammps_metadata_persistence_v1/accelerated_delta.json -text
+
+# Exact result-processing source delta composed on current main.
+tests/fixtures/result_processing_v1/accelerated_delta.json -text
+
+# Exact reviewed PR108 source deltas preserve historical lineage bytes.
+tests/fixtures/pr108_reviewed_source_v1/root_train_policy_delta.json -text
+tests/fixtures/pr108_reviewed_source_v1/easy_train_policy_delta.json -text
+tests/fixtures/pr108_reviewed_source_v1/mechanical_json_scalar_delta.json -text
+
+# Accepted JSON/batch-input source deltas retain exact historical composition.
+tests/fixtures/release_input_contract_v1/*.json -text

fibernet-4.4.0/CHANGELOG.md

@@ -0,0 +1,471 @@
+## [4.4.0] - 2026-11-06
+
+See [4.4.0 release notes and migration](docs/RELEASE_4_4_0.md) for the changes
+since 4.3.0. Historical Unreleased entries below are retained as development
+history and include changes already shipped in earlier releases.
+
+## [Unreleased] - offline LAMMPS saved-run reader
+
+- Add `read_saved_lammps_run` to reconstruct existing saved preparation files
+  and parse captured output without runtime admission or engine execution.
+- Require a caller-supplied manifest digest, exact current-compiler regeneration,
+  bounded POSIX regular-file reads and explicit nonauthentication semantics.
+  Returned SI output is protocol-checked, not certified as original engine data
+  or physically valid; compiler drift is rejected without automatic migration.
+
+## [Unreleased] - gyroid factory parameters
+
+- Forward `lattice_3d(topology="gyroid")` parameters `grid_size`, `radius`, and
+  `material` to the existing backend instead of silently discarding them.
+  Calls without these options now use the documented factory defaults
+  `(3, 3, 3)` and `0.1`, rather than the backend defaults `(2, 2, 2)` and `0.15`.
+  To retain the previous default geometry and radius, pass `grid_size=(2, 2, 2)`
+  and `radius=0.15` explicitly. Direct `gyroid_lattice_3d` defaults are unchanged.
+- This routes parameters to the existing TPMS-inspired fiber-network generator;
+  it does not produce a solid or shell gyroid or qualify mechanical behavior.
+  The existing backend does not use `threshold`, and the factory's `seed` and
+  `perturbation` do not affect this route. Those behaviors are unchanged.
+
+## [Unreleased] - checked graph prediction
+
+- Add `predict_gnn` for existing CPU scalar FiberGNN models and GraphBatch v1,
+  checking caller-declared raw XYZ+dimension compatibility while preserving
+  graph identity and prior module modes. Low-level forward remains unchanged.
+- Return graph-aligned model-space predictions with an explicit caller-supplied
+  target name and unknown target-unit/transform statuses.
+
+## [Unreleased] - raw beam-resultant availability
+
+- Expose the existing component contract on raw 2-D, 3-D and incremental 2-D beam results. Preserve numeric outputs and approximate qualifications; unrecovered 3-D shear slots remain explicitly unavailable.
+- Document caller-qualified field-learning targets and retain unknown legacy component availability.
+
+## [Unreleased] - explicit Taichi geometry policy
+
+- Breaking input acceptance: graph-based spring/Taichi calls now reject nonempty edge-owned `internal_points` by default. Explicit `geometry_policy="endpoint_chord"` retains the existing endpoint interpretation without modifying the graph or claiming curved-fiber mechanics.
+- Forward the policy through direct/facade, streaming batch, Dataset and RL callers; bind batch/Dataset resumes to the policy and helper source. Actual-node straight-edge segmentation remains accepted, and generator/sampling defaults are unchanged.
+- Preserve geometry provenance in ordinary result metadata. Legacy results are not backfilled. See `docs/TAICHI_GEOMETRY_POLICY.md` for migration and scope.
+
+## [Unreleased] - RL evaluation isolation
+
+- Add keyword-only `eval_env` to `fibernet.rl.algorithms.train_rl`, enabling automatic evaluation independently of model saving.
+- Reject shared known in-process training/evaluation environment objects before model or callback construction. Remote resources and custom wrappers still require caller-managed isolation.
+- Save-only calls now warn and skip unsafe automatic evaluation on the training environment, while retaining checkpoint/final saves. Provide an independent `eval_env` to restore best-model selection.
+- Document migration and bounded validation in `docs/architecture/rl-evaluation-isolation.md`; add inert contract regressions without real model training or simulation.
+- Correct root/easy `train_rl` documentation to identify random-policy rollout statistics, not policy learning, and clarify batch damping's existing metadata-only behavior. Runtime and exports are unchanged; source-bound batch resume may require a fresh output directory after this documentation change.
+
+## [Unreleased] - thermal reconstruction budget
+
+- Bound thermal capacity reconstruction with outward-rounded `C*rho`, where `rho` is the current returned-temperature reconstruction radius; retain separate field-evaluation allowances and unchanged algebraic/precision thresholds.
+- Identify this numerical acceptance revision with `thermal_capacity_reconstruction_profile: binary64-rn-capacity-rho-1` in transient and coupled electrothermal provenance. Public call signatures and result layouts are unchanged; numerical bounds and acceptance outcomes may differ.
+- Preserve explicit refusal for lost increments, unrepresentable products and heat scales too small for the precision budget. Smaller time steps are not guaranteed to be representable; this correction does not establish trajectory convergence or coupled-energy accuracy.
+- Include self-contained synthetic acceptance/refusal tests with exact rational equation checks; no external capture file is required.
+
+## [Unreleased] - runtime compatibility
+
+- Use ASCII status labels in easy training and batch helpers so decorative symbols do not break narrow redirected consoles; no locale or stream configuration is changed
+- Read UTF-8 Python source fixtures explicitly in documentation/exporter tests; keep graph JSON loading on its historical default text codec
+- Test invalid UTF-8 batch input with a complete graph and independent UTF-8/cp1252 disk-decoding oracles, separating decoder failures from stub-engine failures
+
+- Temporarily restrict TaichiEngine and accel/full/dev extras to Taichi 1.7.4 for the version-verified runtime-lifetime repair; other versions fail explicitly
+- Same-owner engines share the runtime; owning worker threads are supported, cross-owner use and unverifiable external initialization fail safely
+- Non-Taichi backends are unaffected; broader Taichi versions require explicit lifecycle matrix validation
+
+## [4.3.0] - 2026-10-07
+
+- Add experimental `fibernet.sim.fracture` Config/Solver/Result API and dedicated material-complete JSON CLI
+- Preserve frozen linear Timoshenko, sparse equilibrium and near-tie removal kernels with source-hash provenance
+- Add explicit units/grips, strict geometry/material validation, pre-removal event traces, null free-fragment poses and hash-bound resumable equilibrated checkpoints
+- Add analytical/synthetic regression tests and a runnable public example; retain Python >=3.9 and old `simulate` semantics
+- Add optional `fracture` dependencies (psutil, threadpoolctl); do not bundle LAMMPS or private research inputs/results
+- Replace hard-coded 4.2.0 PyPI CI check with an explicit, version-parameterized post-publication verification workflow
+- Scope: uncalibrated small-displacement static software; no automatic contact, finite-separation proof or fracture-toughness claim
+
+## [4.1.4] - 2026-07-23
+
+### Changed
+- Rename `BeamFrameFEM_v6` → `BeamFrameFEM` (public API alias; `BeamFrameFEM_v6` still available for backward compatibility)
+- Split Simulation section into Mass-Spring and FEM subsections with comparison table (physics, joints, radius effect, speed, use cases)
+- Reorder showcase images: generation → mass-spring → trajectory → FEM → ML → RL
+- Remove Performance section from README
+
+### Added
+- `BeamFrameFEM` lazy import in `fibernet.ml` (importable as `from fibernet.ml import BeamFrameFEM`)
+- FEM API section with linear/nonlinear solver explanation
+- Mass-Spring vs FEM comparison table in both README.md and README_CN.md
+
+## [4.1.3] - 2026-07-23
+
+### Documentation
+- Bilingual README overhaul: added dedicated Beam Frame FEM section with showcase images, API examples, and verification results
+- README.md and README_CN.md updated to v4.1.2 feature set (26 unit types, dual simulation, FEM physics model)
+- Added FEM showcase images (dark + light themes) to `docs/images/`
+
+### Improved
+- Reorganized root-level files: community health files (CONTRIBUTING, SECURITY, CODE_OF_CONDUCT) moved to `.github/`, ATTRIBUTIONS.md moved to `docs/`, utility scripts moved to `scripts/`
+- Updated MANIFEST.in for new file locations
+
+## [4.1.2] - 2026-07-23
+
+### Fixed
+- **Critical**: `graph_to_fem_input()` now correctly extracts per-edge radius from StructureGraph
+  - Root cause: iterating `graph.edges` (Dict) gave keys (ints) instead of SEdge values
+  - FEM results now properly reflect fiber diameter changes (welded joint physics)
+- `to_sim_result()` radius access fixed for energy calculation
+
+### Improved
+- FEM showcase: restructured layout (5 rows) fixing middle-row panel overflow
+- Dark theme: increased edge linewidth (0.6→2.0), switched to brighter colormaps (magma/RdYlBu)
+- Radius variation row now shows physical stress dependence on fiber diameter
+
+## [4.1.1] - 2026-07-23
+
+### Added
+- `BeamFrameFEM_v6.graph_to_fem_input()`: StructureGraph to FEM input conversion
+- `BeamFrameFEM_v6.stretch_test()`: One-liner uniaxial FEM stretch test
+- `BeamFrameFEM_v6.to_sim_result()`: Convert FEM dict to SimResult
+- `render_fem_stress()`: FEM stress visualization (dark/light themes)
+
+### Fixed
+- solve_2d and solve_3d: Invalid node index now raises clear ValueError
+- solve_3d: Now returns reactions, edge_forces, K matrix (matching solve_2d)
+
+### API Improvement
+- Basic FEM analysis: ~10 lines reduced to 3 lines
+
+## [4.1.0] - 2026-07-23
+
+### Added
+- **BeamFrameFEM_v6**: Production-grade beam frame finite element method
+  - `solve_2d()`: Linear 2D beam frame analysis with axial + bending + shear
+  - `solve_2d_nonlinear()`: Geometrically nonlinear solver (co-rotational incremental)
+  - `solve_3d()`: 3D beam frame analysis
+  - Full stress decomposition: axial, bending, total stresses per element
+  - Reaction force computation at fixed boundaries
+  - Moment distribution along beams
+  - Validated against analytical cantilever, displacement BC, and nonlinear benchmarks
+
+- **Large Deformation Test Suite**: Comprehensive FEM validation framework
+  - 152 simulations across 14 structure types (8 2D + 6 3D)
+  - 4 fiber radii (0.02, 0.05, 0.10, 0.20)
+  - Stretch targets: 2.0x, 1.5x, 0.7x, 0.5x
+  - Deformation propagation analysis (boundary vs interior displacement)
+  - Bending vs axial dominance quantification
+  - Checkpoint/resume support for long-running tests
+  - Key finding: 100% FULL deformation propagation, all structures BENDING-dominated
+
+- **14 3D unit types** (up from initial set):
+  `bcc`, `chiral_3d`, `cubic`, `diamond_3d`, `fcc`, `gyroid`, `hcp`, `iwp`,
+  `lidinoid`, `neovius`, `octet`, `reentrant_3d`, `schwarz_d`, `schwarz_p`
+
+- **12 2D unit types**: `chiral`, `cross`, `diamond`, `hexagon`, `honeycomb`,
+  `kagome`, `missing_rib`, `reentrant`, `square`, `star`, `triangle`, `voronoi`
+
+### Fixed
+- `easy.py`: `simulate()` default mode changed from `"tension"` (invalid) to `"stretch"` (correct)
+- `easy.py`: Docstring examples updated to match actual API
+
+### Changed
+- Version bumped to 4.1.0 (minor release: new FEM module + 3D structures)
+
+## [4.0.5] - 2026-07-17
+
+### Fixed
+- Vectorized `compute_detailed()` edge length computation (same pattern as trajectory save)
+
+### Changed
+- `dynamics()` now supports `num_steps=0` for kernel warmup (field allocation + kernel compilation without simulation)
+- Notebook (Cell 19) updated: kernel warmup added, JSON save restored with tqdm progress
+
+## [4.0.4] - 2026-07-17
+
+### Fixed
+- **Critical: Progressive slowdown during batch simulation (kernel recompilation)**
+  - Moved `@ti.kernel def substep()` out of `dynamics()` method body
+  - Kernel now compiled once per `(dim, num_nodes, num_edges)` and cached alongside fields
+  - Before: Each `dynamics()` call recompiled the kernel → Taichi LLVM accumulated modules → progressive slowdown
+  - After: Kernel cached → constant per-structure time across hundreds of calls
+  - Test: 50 structures, first-10 avg = 6.60s, last-10 avg = 7.09s (ratio 1.07x — no slowdown)
+
+### Technical Details
+- File: `fibernet/sim/accelerated.py`
+- Change: Kernel definition moved inside the `_field_cache` allocation block, stored as `cached['substep']`
+- Impact: Eliminates ~0.5-1s kernel compilation overhead per call after the first
+- Combined with 4.0.2 (vectorized save) and 4.0.3 (field cache): total ~16x speedup from original
+
+## [4.0.3] - 2026-07-17
+
+### Fixed
+- **Critical: Taichi SNode exhaustion causing simulation hang at ~128 structures**
+  - Added `_field_cache` to `TaichiEngine` class
+  - Reuses Taichi fields across calls with same `(dim, num_nodes, num_edges)`
+  - Before: Each `dynamics()` call allocated 14 Taichi fields → SNode limit (~1024) hit at ~128 calls → process hangs indefinitely
+  - After: Fields cached and reused → tested 150+ structures without hang
+  - Memory usage reduced significantly (no field accumulation)
+
+### Technical Details
+- File: `fibernet/sim/accelerated.py`
+- Change: Added class-level `_field_cache` dict keyed by `(dim, num_nodes, num_edges)`
+- Impact: Fixes critical hang bug, improves memory efficiency
+- Test: Successfully ran 150 structures sequentially (previously hung at #128)
+
+## [4.0.2] - 2026-07-16
+
+### Performance
+- **16x speedup** in simulation trajectory saving
+  - Vectorized edge length computation in `dynamics()` method
+  - Replaced Python list comprehension with NumPy vectorized operation
+  - Before: ~93s per structure (30000 steps, save_interval=500)
+  - After: ~27s per structure with same parameters
+  - With recommended parameters (8000 steps): ~5.8s per structure
+
+### Fixed
+- Simulation appearing to "hang" with many structures
+  - Root cause: excessive overhead in trajectory save loop
+  - Each save took ~1.17s due to Python loop over all edges
+  - Now <1ms per save with vectorized computation
+
+### Technical Details
+- File: `fibernet/sim/accelerated.py`
+- Change: `np.array([np.linalg.norm(...) for e in range(n)])` → `np.linalg.norm(..., axis=1)`
+- Impact: Pure performance optimization, no API or numerical changes
+- All 118 existing tests pass
+
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [1.18.0] - 2026-07-05
+
+### Added
+- **Mesh export module** (`fibernet/io/mesh_export.py`)
+  - Export to STL (ASCII and binary)
+  - Export to Wavefront OBJ
+  - Export to Stanford PLY
+  - Cylindrical mesh generation for fibers
+  - Configurable polygon approximation (n_sides)
+
+### Tests
+- 9 new tests in `tests/test_mesh_export.py`
+- Total: 752 passing, 6 skipped
+
+## [1.17.0] - 2026-07-05
+
+### Added
+- **Network comparison module** (`fibernet/analysis/comparison.py`)
+  - `NetworkFingerprint`: structural fingerprint computation
+  - `NetworkComparator`: pairwise distances, clustering, similarity search
+  - `compare_networks()`: compare multiple networks
+  - `network_similarity()`: pairwise similarity score
+- Example 14: advanced structural analysis workflow
+
+### Tests
+- 14 new tests in `tests/test_comparison.py`
+- Total: 743 passing, 6 skipped
+
+## [1.16.0] - 2026-07-05
+
+### Added
+- **Homogenization module** (`fibernet/analysis/homogenization.py`)
+  - `EffectiveElasticProperties`: E, nu, G computation (2D/3D)
+  - `EffectiveThermalProperties`: thermal conductivity, CTE
+  - `EffectiveElectricalProperties`: electrical conductivity
+  - `compute_effective_properties()`: comprehensive analysis
+
+### Tests
+- 13 new tests in `tests/test_homogenization.py`
+- Total: 729 passing, 6 skipped
+
+## [1.15.0] - 2026-07-05
+
+### Added
+- **Spatial statistics module** (`fibernet/analysis/spatial.py`)
+  - `SpatialStatistics`: Ripley's K, pair correlation, nearest neighbor
+  - `OrientationAnalysis`: nematic order, orientation histograms
+  - `LengthAnalysis`: length distributions, statistics, fitting
+  - `ConnectivityAnalysis`: degree distribution, mean connectivity
+  - `AnisotropyAnalysis`: fabric tensor, anisotropy index
+  - `compute_spatial_statistics()`: comprehensive analysis
+
+### Tests
+- 22 new tests in `tests/test_advanced_statistics.py`
+- Total: 716 passing, 6 skipped
+
+## [1.14.0] - 2026-07-05
+
+### Added
+- **Material database** (`fibernet/materials.py`)
+  - 29 pre-defined materials (polymers, metals, ceramics, biological, carbon)
+  - `get_material()`: quick material lookup
+  - `list_materials()`: list all available materials
+  - `compare_materials()`: material selection tool
+  - `m()`: shortcut for get_material
+- **Unit conversion utilities** (`fibernet/units.py`)
+  - Length: m, mm, um, nm, cm, km, in, ft
+  - Force: N, kN, mN, µN, nN, lbf, dyne
+  - Pressure: Pa, kPa, MPa, GPa, TPa, bar, atm, psi, ksi
+  - Temperature: K, C, F, R
+  - Energy: J, kJ, mJ, cal, kcal, eV, BTU, kWh
+
+### Tests
+- 33 new tests in `tests/test_materials.py` and `tests/test_units.py`
+- Total: 694 passing, 6 skipped
+
+## [1.13.0] - 2026-07-05
+
+### Added
+- **Fractal network generators** (`fibernet/gen/fractal.py`)
+  - Sierpinski triangle
+  - Koch curve
+  - Fractal tree
+  - Hilbert curve
+- **Gradient network generators** (`fibernet/gen/gradient.py`)
+  - Density gradient (linear, exponential, sinusoidal)
+  - Property gradient (material properties varying spatially)
+  - Multi-zone networks (custom regions with different properties)
+- Example 13: publication-ready research workflow
+
+### Tests
+- 29 new tests in `tests/test_fractal.py` and `tests/test_gradient.py`
+- Total: 687 passing, 6 skipped
+
+## [1.12.0] - 2026-07-04
+
+### Added
+- Network topology analysis module
+- Comprehensive topology metrics
+
+### Tests
+- Total: 658 passing
+
+## [1.11.0] - 2026-07-04
+
+### Added
+- Visualization module (2D/3D plotting)
+- Design of Experiments (DOE) module
+  - Grid search
+  - Latin hypercube sampling
+  - Random sampling
+  - Sensitivity analysis
+
+### Tests
+- Total: 635 passing
+
+## [1.10.0] - 2026-07-04
+
+### Added
+- Fatigue analysis module
+- Creep analysis module
+- Diffusion/transport module
+
+### Tests
+- Total: 607 passing
+
+## [1.9.0] - 2026-07-03
+
+### Added
+- Benchmark suite for performance testing
+- Enhanced multiscale testing
+
+### Tests
+- Total: 585 passing
+
+## [1.8.0] - 2026-07-03
+
+### Added
+- Electromagnetic analysis
+- Molecular dynamics integration
+- Open-source tool integrations
+- Research case study examples
+
+### Tests
+- Total: 575 passing
+
+## [1.7.0] - 2026-07-02
+
+### Added
+- Graph neural network (GNN) module
+- Permeability analysis
+- Uncertainty quantification (UQ)
+- Coefficient of thermal expansion (CTE)
+
+### Tests
+- Total: 544 passing
+
+## [1.6.0] - 2026-07-02
+
+### Added
+- Incremental FEM solver
+- Buckling analysis module
+
+### Tests
+- Total: 477 passing
+
+## [1.5.0] - 2026-07-01
+
+### Added
+- Nonlinear mechanics module
+- Damage mechanics
+- Viscoelasticity
+
+### Tests
+- Total: 448 passing
+
+## [1.4.0] - 2026-07-01
+
+### Added
+- Periodic boundary conditions (PBC)
+- Transforms module
+- Advanced generators (chiral, woven, hierarchical)
+
+### Tests
+- Total: 410 passing
+
+## [1.3.0] - 2026-06-30
+
+### Added
+- Machine learning module
+  - Property prediction
+  - Feature engineering
+- Percolation analysis
+- Integration tests
+
+### Tests
+- Total: 370 passing
+
+## [1.2.0] - 2026-06-30
+
+### Added
+- Taichi-accelerated FEM solver
+- I/O modules (VTK, LAMMPS, GMSH, PDB, XYZ)
+- Pandas integration
+
+### Tests
+- Total: 320 passing
+
+## [1.1.0] - 2026-06-29
+
+### Added
+- Simulation modules (thermal, fracture, etc.)
+- Core data structures
+- Basic generators
+
+### Tests
+- Total: 250 passing
+
+## [1.0.0] - 2026-06-28
+
+### Added
+- Initial release
+- Core fiber network data structures
+- Basic random network generators
+- Simple FEM solver
+- Sphinx documentation framework
+- GitHub Actions CI/CD
+- pyproject.toml for pip installation
+
+### Tests
+- Total: 180 passing
+

fibernet-4.4.0/FiberScope/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 FiberNet Contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

fibernet-4.4.0/FiberScope/THIRD_PARTY_NOTICES.md

@@ -0,0 +1,11 @@
+# Third-party and asset notices
+
+FiberScope application source and the FiberNet toolkit are distributed under the repository MIT license. Third-party libraries retain their own licenses. License texts collected for the desktop distribution are under assets/licenses; optional training dependencies are not all embedded in the EXE.
+
+PySide6 / Qt for Python and Shiboken are used under their applicable LGPL/GPL/commercial terms, with the relevant license texts included. The application source and Windows build script are supplied to permit rebuilding with compatible replacement libraries. Third-party software is not relicensed under MIT by this repository.
+
+## Geometry assets
+
+demo_pyramid, demo_scarf, demo_sleeve, demo_tshirt and demo_vest are procedural demonstration geometries generated by project code. Reference-derived shirt, shoe and tower envelopes have transformation details and hashes in assets/obj/reference26.json. The author identifies the displayed products as transformed fibrous-network structures. Upstream redistribution licenses for those reference shapes and the older lung/heart/shoe assets have not been supplied; this notice does not represent those source shapes as MIT-original works or grant additional upstream rights. Original reference documents and high-resolution source OBJ files are not part of this release. The public web demo uses only procedural pyramid/sleeve/scarf models.
+
+Bambu Studio is an external application. It is not bundled, and no printer credentials are included.