PyPI - loadpath - Versions diffs - 0.1.1__tar.gz → 0.1.3__tar.gz - Mend

loadpath 0.1.1tar.gz → 0.1.3tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (297) hide show

{loadpath-0.1.1 → loadpath-0.1.3}/PKG-INFO RENAMED Viewed

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: loadpath
-Version: 0.1.1
+Version: 0.1.3
 Summary: Generative design for printed parts: topology optimization from keep-outs, preserved geometry and loads, verified in real units.
 License-Expression: MIT
 Project-URL: Homepage, https://github.com/connorwhite-online/loadpath
@@ -70,18 +70,19 @@ safety factor, deflection, thinnest member, printability, pass or fail.
 <table>
 <tr>
-<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/printed-hook.png" width="380"></td>
-<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/stress/printed-hook.png" width="380"></td>
+<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/grip-wrist-link-mjf-part.png" height="420"></td>
+<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/stress/grip-wrist-link-mjf.png" height="420"></td>
 </tr>
 <tr>
-<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/bell-crank.png" width="380"></td>
-<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/stress/bell-crank.png" width="380"></td>
+<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/grip-wrist-yoke-part.png" width="380"></td>
+<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/stress/grip-wrist-yoke.png" width="380"></td>
 </tr>
 <tr>
-<td align="center" colspan="2"><sub>Left, the part loadpath delivered (blue: load, red: kept bore). Right, its
-stress from a fresh analysis of that mesh, blue (none) to red (the part's 98th percentile), on the shape bent by
-the worst load case. Top: a 300 N wall hook printed upright without supports, 24.2 g, SF 2.73. Bottom: a 500 N
-aluminium bell crank (LPBF), 23.4 g, SF 3.22. More in <a href="#examples">Examples</a>.</sub></td>
+<td align="center" colspan="2"><sub>Two robot-arm parts as loadpath delivered them (left) and their stress from a fresh
+analysis of that mesh (right), blue (none) to red (the part's 98th percentile), on the shape bent by the worst
+load case. Top: a wrist link between two servos, MJF nylon, 40.2 g, safety factor 5.8. Bottom: a servo cradle
+held only at its eight screw bosses, glass-filled SLS nylon, 17.3 g, safety factor 3.9. More in
+<a href="#examples">Examples</a>.</sub></td>
 </tr>
 </table>
@@ -99,12 +100,46 @@ Then ask in plain words:
 > printed on an FDM printer.
 The agent writes the study, checks it, runs it in the background and reads
-the report back. Its guide is [loadpath/SKILL.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/loadpath/SKILL.md) (the
+the report back. Its guide is [loadpath/SKILL.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/loadpath/SKILL.md) (the
 `loadpath_guide` tool returns it); the tools and the loop are in
-[docs/agents.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/agents.md). The server runs locally and nothing
+[docs/agents.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/agents.md). The server runs locally and nothing
 leaves the machine. Units are millimetres, newtons and megapascals. MIT
 licensed.
+## Where it fits
+A designed part needs a shape, the loads it carries, and a way to make it.
+The open tools each cover one of those:
+- **Topology optimization codes** (top88, top3d, ToPy, FEniCS scripts) are
+  research solvers. They stop at a density field on a grid: no printable
+  part, and nothing checks the result.
+- **Script CAD** (CadQuery, build123d, OpenSCAD) and AI CAD tools draw exact
+  geometry but don't know the loads.
+- **Commercial generative design** (Fusion, nTop) does the whole job, but
+  it's built around a GUI, not an agent.
+loadpath is the layer in between. Your CAD defines the interfaces; loadpath
+takes those plus the loads, material and process, and returns a watertight
+part checked by an analysis independent of the optimizer, with a JSON report
+an agent can act on. Its STEP output goes back into CAD.
+Where it is strongest:
+- **The analysis behind the shape**: buckling, fatigue and creep,
+  large-deflection compliant mechanisms, loads of unknown direction, thermal
+  loads and print warp, all checked on the delivered mesh.
+- **Implicit modelling**: field graphs, graded and conformal lattices (TPMS,
+  strut, Voronoi), blends, and exact STEP of lattice parts.
+- **Process awareness**: FDM, SLS, MJF, resin and powder-bed metal rules,
+  plus casting and milling constraints.
+- **Heat exchangers**: lattice-Boltzmann flow and conjugate heat transfer
+  for gyroid cores and cold plates.
+Not yet: draft-angle and machining-time checks for cast and milled parts,
+or a ranked spread of outcomes across materials and processes by cost.
+There is no GUI on purpose; it's a library.
 ## What it does
 - **Topology optimization** on a voxel grid (SIMP with a density filter,
@@ -121,23 +156,23 @@ licensed.
   clean-outs), plus milling and casting constraints.
 - **Field modelling**: a part as a graph of implicit fields, so skins,
   lattices (TPMS and strut) and fillets can be graded by stress or by
-  distance ([docs/fields.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/fields.md)).
+  distance ([docs/fields.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/fields.md)).
 - **Other analyses**: large deflection and buckling, compliant mechanisms,
   thermal loads, fatigue and creep estimates, lattice crush and impact,
   vibration, lattice-Boltzmann flow and conjugate heat for heat exchangers
   and cold plates, and layer-by-layer warp and residual stress
-  ([docs/BUILDSIM.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/BUILDSIM.md)).
+  ([docs/BUILDSIM.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/BUILDSIM.md)).
 - **CAD in and out**: a STEP or STL part becomes a draft study with its
-  bores and bolt patterns found ([docs/IMPORT.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/IMPORT.md)); results
+  bores and bolt patterns found ([docs/IMPORT.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/IMPORT.md)); results
   export as STL, 3MF, and STEP with exact planes and cylinders at the
   interfaces; a lattice part (strut cells, TPMS, graded, skinned or open)
-  goes to STEP lattice and all, as smooth patches on its true surface ([docs/fields.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/fields.md#the-lattice-itself-in-the-step-lattice-exact)).
+  goes to STEP lattice and all, as smooth patches on its true surface ([docs/fields.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/fields.md#the-lattice-itself-in-the-step-lattice-exact)).
 - **Mesh checks without a study** (`loadpath.checks`): wall thickness,
   overhangs, trapped voids, fluid-network isolation, component fit.
 ## How much to trust it
-Read [docs/VALIDATION.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/VALIDATION.md) before relying on a number.
+Read [docs/VALIDATION.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/VALIDATION.md) before relying on a number.
 In short:
 - **The solvers match analytic answers** in the test suite: beam
@@ -162,16 +197,16 @@ In short:
 ## Case studies
 Each states the problem, the setup, the numbers against a named baseline,
-and the limitations ([docs/case-studies/](https://github.com/connorwhite-online/loadpath/tree/v0.1.1/docs/case-studies/)).
+and the limitations ([docs/case-studies/](https://github.com/connorwhite-online/loadpath/tree/v0.1.3/docs/case-studies/)).
 | Case study | Result | Baseline |
 | --- | --- | --- |
-| [Eval suite](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/case-studies/evals.md) | 15 of 15 parts meet their requirements on a fresh analysis of the delivered mesh | Each part's own requirements |
-| [GRIP robot arm](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/case-studies/grip.md) | A load sweep over the arm's URDF found 3 of 5 links below SF 1; the regrown set meets SF 2 and is 18% smaller | GRIP's committed parts, made by an earlier loadpath with hand-written loads |
-| [Arm link cores](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/case-studies/arm-link-cores.md) | 42.2 g box section for FDM, 36.8 g hollow link for SLS, both at 0.5 mm and SF 3 | 54.9 g solid web; 54.2 g free-form topology optimization |
-| [Liquid cold plate](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/case-studies/cold-plate.md) | A die-graded gyroid ties streamline channels; channels filled with a graded gyroid run 7% cooler | Streamline channels and pin fins, equal metal and pressure drop |
-| [TPU finger](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/case-studies/tpu-finger.md) | up to 8 mm of tip curl for a 5 mm pull; a tendon finger grips 8x harder | A linear-model design (2.5 mm, then the solve fails); a plain bar of equal volume |
-| [Benchmarks](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/case-studies/benchmarks.md) | 1-5% less stiff than top88/top3d at equal material, in 57-60 iterations instead of 130-410 | Ports of top88 and top3d |
+| [Eval suite](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/case-studies/evals.md) | 15 of 15 parts meet their requirements on a fresh analysis of the delivered mesh | Each part's own requirements |
+| [GRIP robot arm](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/case-studies/grip.md) | A load sweep over the arm's URDF found 3 of 5 links below SF 1; the regrown set meets SF 2 and is 18% smaller | GRIP's committed parts, made by an earlier loadpath with hand-written loads |
+| [Arm link cores](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/case-studies/arm-link-cores.md) | 42.2 g box section for FDM, 36.8 g hollow link for SLS, both at 0.5 mm and SF 3 | 54.9 g solid web; 54.2 g free-form topology optimization |
+| [Liquid cold plate](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/case-studies/cold-plate.md) | A die-graded gyroid ties streamline channels; channels filled with a graded gyroid run 7% cooler | Streamline channels and pin fins, equal metal and pressure drop |
+| [TPU finger](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/case-studies/tpu-finger.md) | up to 8 mm of tip curl for a 5 mm pull; a tendon finger grips 8x harder | A linear-model design (2.5 mm, then the solve fails); a plain bar of equal volume |
+| [Benchmarks](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/case-studies/benchmarks.md) | 1-5% less stiff than top88/top3d at equal material, in 57-60 iterations instead of 130-410 | Ports of top88 and top3d |
 ## Command line and Python
@@ -190,7 +225,7 @@ numba), `gpu` (PyTorch for large grids on CUDA, Apple Silicon or CPU
 threads).
 Studies are validated against a published JSON Schema
-([`schema/`](https://github.com/connorwhite-online/loadpath/tree/v0.1.1/schema/)), and `loadpath check` returns every problem at once
+([`schema/`](https://github.com/connorwhite-online/loadpath/tree/v0.1.3/schema/)), and `loadpath check` returns every problem at once
 with its JSON path, so an agent can fix a study in one round trip. Long
 solves run as background jobs the agent polls. Each finished outcome has an
 `attention` list of numbers to look at before trusting it, and renders a
@@ -210,51 +245,51 @@ print(best.report["mass"], best.report["safetyFactor"])
 ## Documentation
-- [Quickstart](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/quickstart.md): install to a checked part.
-- [Tutorials](https://github.com/connorwhite-online/loadpath/tree/v0.1.1/docs/tutorials/): eleven end-to-end workflows, each run by CI
+- [Quickstart](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/quickstart.md): install to a checked part.
+- [Tutorials](https://github.com/connorwhite-online/loadpath/tree/v0.1.3/docs/tutorials/): eleven end-to-end workflows, each run by CI
   (an FDM bracket to 3MF, an SLS lattice with drain holes, a metal heat
   exchanger, an impact bumper, a STEP part round trip, LPBF distortion
   compensation, and more).
-- [Reference](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/loadpath/README.md): the method, every study field, the
+- [Reference](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/loadpath/README.md): the method, every study field, the
   report, the limits.
-- [Validation](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/VALIDATION.md): what each model has been checked
+- [Validation](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/VALIDATION.md): what each model has been checked
   against.
-- [Case studies](https://github.com/connorwhite-online/loadpath/tree/v0.1.1/docs/case-studies/) and [benchmarks](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/BENCHMARKS.md).
-- Topics: [fields](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/fields.md), [materials](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/materials.md),
-  [build simulation](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/BUILDSIM.md), [calibration](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/CALIBRATION.md),
-  [CAD import](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/IMPORT.md), [slicer](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/slicer.md),
-  [multi-material](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/MULTIMATERIAL.md), [impact](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/IMPACT.md),
-  [mechanisms](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/MECHANISMS.md).
+- [Case studies](https://github.com/connorwhite-online/loadpath/tree/v0.1.3/docs/case-studies/) and [benchmarks](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/BENCHMARKS.md).
+- Topics: [fields](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/fields.md), [materials](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/materials.md),
+  [build simulation](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/BUILDSIM.md), [calibration](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/CALIBRATION.md),
+  [CAD import](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/IMPORT.md), [slicer](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/slicer.md),
+  [multi-material](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/MULTIMATERIAL.md), [impact](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/IMPACT.md),
+  [mechanisms](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/MECHANISMS.md).
 <!-- gallery:start -->
 ## Examples
-The eval suite (15/15 passing): robot-arm links, brackets, a bell crank and a hook, each optimized and then checked by a fresh analysis of the delivered mesh. "Under envelope" compares with the design space filled solid, not with a conventionally designed part. These are simulated results; see [docs/VALIDATION.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/VALIDATION.md). Blue arrows are loads, red is keep-out. Every view, stress map and number is in [docs/GALLERY.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/GALLERY.md).
+The eval suite (15/15 passing): robot-arm links, brackets, a bell crank and a hook, each optimized and then checked by a fresh analysis of the delivered mesh. "Under envelope" compares with the design space filled solid, not with a conventionally designed part. These are simulated results; see [docs/VALIDATION.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/VALIDATION.md). Blue arrows are loads, red is keep-out. Every view, stress map and number is in [docs/GALLERY.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/GALLERY.md).
 <table>
 <tr>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/arm-link.png" width="250"><br><b>arm-link</b><br><sub>55.3 g · 56% under envelope · SF 8.92 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/arm-link-cast.png" width="250"><br><b>arm-link-cast</b><br><sub>67.8 g · 46% under envelope · SF 20.11 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/arm-link-driven.png" width="250"><br><b>arm-link-driven</b><br><sub>48.4 g · 61% under envelope · SF 11.17 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/arm-link.png" width="250"><br><b>arm-link</b><br><sub>55.3 g · 56% under envelope · SF 8.92 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/arm-link-cast.png" width="250"><br><b>arm-link-cast</b><br><sub>67.8 g · 46% under envelope · SF 20.11 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/arm-link-driven.png" width="250"><br><b>arm-link-driven</b><br><sub>48.4 g · 61% under envelope · SF 11.17 · pass</sub></td>
 </tr>
 <tr>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/bell-crank.png" width="250"><br><b>bell-crank</b><br><sub>23.4 g · 84% under envelope · SF 3.22 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/grip-turret.png" width="250"><br><b>grip-turret</b><br><sub>140.4 g · 91% under envelope · SF 2.85 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/grip-upper-arm.png" width="250"><br><b>grip-upper-arm</b><br><sub>207.7 g · 88% under envelope · SF 7.33 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/bell-crank.png" width="250"><br><b>bell-crank</b><br><sub>23.4 g · 84% under envelope · SF 3.22 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/grip-turret.png" width="250"><br><b>grip-turret</b><br><sub>140.4 g · 91% under envelope · SF 2.85 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/grip-upper-arm.png" width="250"><br><b>grip-upper-arm</b><br><sub>207.7 g · 88% under envelope · SF 7.33 · pass</sub></td>
 </tr>
 <tr>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/grip-wrist-link-mjf.png" width="250"><br><b>grip-wrist-link-mjf</b><br><sub>40.2 g · 95% under envelope · SF 5.83 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/grip-wrist-yoke.png" width="250"><br><b>grip-wrist-yoke</b><br><sub>17.3 g · 93% under envelope · SF 3.88 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/mbb-beam.png" width="250"><br><b>mbb-beam</b><br><sub>13.8 g · 79% under envelope · SF 1.51 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/grip-wrist-link-mjf.png" width="250"><br><b>grip-wrist-link-mjf</b><br><sub>40.2 g · 95% under envelope · SF 5.83 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/grip-wrist-yoke.png" width="250"><br><b>grip-wrist-yoke</b><br><sub>17.3 g · 93% under envelope · SF 3.88 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/mbb-beam.png" width="250"><br><b>mbb-beam</b><br><sub>13.8 g · 79% under envelope · SF 1.51 · pass</sub></td>
 </tr>
 <tr>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/motor-mount.png" width="250"><br><b>motor-mount</b><br><sub>27.6 g · 69% under envelope · SF 3.65 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/printed-hook.png" width="250"><br><b>printed-hook</b><br><sub>24.2 g · 84% under envelope · SF 2.73 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/wall-bracket.png" width="250"><br><b>wall-bracket</b><br><sub>26.9 g · 94% under envelope · SF 2.85 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/motor-mount.png" width="250"><br><b>motor-mount</b><br><sub>27.6 g · 69% under envelope · SF 3.65 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/printed-hook.png" width="250"><br><b>printed-hook</b><br><sub>24.2 g · 84% under envelope · SF 2.73 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/wall-bracket.png" width="250"><br><b>wall-bracket</b><br><sub>26.9 g · 94% under envelope · SF 2.85 · pass</sub></td>
 </tr>
 <tr>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/wall-bracket-fasteners.png" width="250"><br><b>wall-bracket-fasteners</b><br><sub>26.5 g · 94% under envelope · SF 2.57 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/wall-bracket-fasteners.png" width="250"><br><b>wall-bracket-fasteners</b><br><sub>26.5 g · 94% under envelope · SF 2.57 · pass</sub></td>
 </tr>
 </table>
@@ -262,23 +297,23 @@ Where the load goes: von Mises stress, blue (none) to red (the part's 98th perce
 <table>
 <tr>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/stress/bell-crank.png" width="250"><br><sub>bell-crank</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/stress/grip-upper-arm.png" width="250"><br><sub>grip-upper-arm</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/stress/printed-hook.png" width="250"><br><sub>printed-hook</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/stress/bell-crank.png" width="250"><br><sub>bell-crank</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/stress/grip-upper-arm.png" width="250"><br><sub>grip-upper-arm</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/stress/printed-hook.png" width="250"><br><sub>printed-hook</sub></td>
 </tr>
 </table>
 Two-fluid heat exchangers: two isolated channel networks through one sheet gyroid, with headers that print without support.
-![box-crossflow](https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/heatx/box-crossflow.png)
+![box-crossflow](https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/heatx/box-crossflow.png)
 Printed in metal, nothing inside can be supported: the channels end in peaked roofs instead of flat ceilings (red).
-![metal ceilings](https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/heatx/lpbf-ceilings.png)
+![metal ceilings](https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/heatx/lpbf-ceilings.png)
 Made another way: the same arm link printed, cast (no undercuts along z), 3-axis milled from both sides with a 6 mm cutter, and with members capped at 6 mm. Each meets SF 3 and 0.5 mm.
-![manufacturing constraints](https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/manufacturing/arm-link.png)
+![manufacturing constraints](https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/manufacturing/arm-link.png)
 <sub>Rendered 2026-11-03 from `6c97b57` by `python docs/gallery.py`.</sub>

{loadpath-0.1.1 → loadpath-0.1.3}/README.md RENAMED Viewed

@@ -9,18 +9,19 @@ safety factor, deflection, thinnest member, printability, pass or fail.
 <table>
 <tr>
-<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/printed-hook.png" width="380"></td>
-<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/stress/printed-hook.png" width="380"></td>
+<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/grip-wrist-link-mjf-part.png" height="420"></td>
+<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/stress/grip-wrist-link-mjf.png" height="420"></td>
 </tr>
 <tr>
-<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/bell-crank.png" width="380"></td>
-<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/stress/bell-crank.png" width="380"></td>
+<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/grip-wrist-yoke-part.png" width="380"></td>
+<td align="center" width="50%"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/stress/grip-wrist-yoke.png" width="380"></td>
 </tr>
 <tr>
-<td align="center" colspan="2"><sub>Left, the part loadpath delivered (blue: load, red: kept bore). Right, its
-stress from a fresh analysis of that mesh, blue (none) to red (the part's 98th percentile), on the shape bent by
-the worst load case. Top: a 300 N wall hook printed upright without supports, 24.2 g, SF 2.73. Bottom: a 500 N
-aluminium bell crank (LPBF), 23.4 g, SF 3.22. More in <a href="#examples">Examples</a>.</sub></td>
+<td align="center" colspan="2"><sub>Two robot-arm parts as loadpath delivered them (left) and their stress from a fresh
+analysis of that mesh (right), blue (none) to red (the part's 98th percentile), on the shape bent by the worst
+load case. Top: a wrist link between two servos, MJF nylon, 40.2 g, safety factor 5.8. Bottom: a servo cradle
+held only at its eight screw bosses, glass-filled SLS nylon, 17.3 g, safety factor 3.9. More in
+<a href="#examples">Examples</a>.</sub></td>
 </tr>
 </table>
@@ -38,12 +39,46 @@ Then ask in plain words:
 > printed on an FDM printer.
 The agent writes the study, checks it, runs it in the background and reads
-the report back. Its guide is [loadpath/SKILL.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/loadpath/SKILL.md) (the
+the report back. Its guide is [loadpath/SKILL.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/loadpath/SKILL.md) (the
 `loadpath_guide` tool returns it); the tools and the loop are in
-[docs/agents.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/agents.md). The server runs locally and nothing
+[docs/agents.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/agents.md). The server runs locally and nothing
 leaves the machine. Units are millimetres, newtons and megapascals. MIT
 licensed.
+## Where it fits
+A designed part needs a shape, the loads it carries, and a way to make it.
+The open tools each cover one of those:
+- **Topology optimization codes** (top88, top3d, ToPy, FEniCS scripts) are
+  research solvers. They stop at a density field on a grid: no printable
+  part, and nothing checks the result.
+- **Script CAD** (CadQuery, build123d, OpenSCAD) and AI CAD tools draw exact
+  geometry but don't know the loads.
+- **Commercial generative design** (Fusion, nTop) does the whole job, but
+  it's built around a GUI, not an agent.
+loadpath is the layer in between. Your CAD defines the interfaces; loadpath
+takes those plus the loads, material and process, and returns a watertight
+part checked by an analysis independent of the optimizer, with a JSON report
+an agent can act on. Its STEP output goes back into CAD.
+Where it is strongest:
+- **The analysis behind the shape**: buckling, fatigue and creep,
+  large-deflection compliant mechanisms, loads of unknown direction, thermal
+  loads and print warp, all checked on the delivered mesh.
+- **Implicit modelling**: field graphs, graded and conformal lattices (TPMS,
+  strut, Voronoi), blends, and exact STEP of lattice parts.
+- **Process awareness**: FDM, SLS, MJF, resin and powder-bed metal rules,
+  plus casting and milling constraints.
+- **Heat exchangers**: lattice-Boltzmann flow and conjugate heat transfer
+  for gyroid cores and cold plates.
+Not yet: draft-angle and machining-time checks for cast and milled parts,
+or a ranked spread of outcomes across materials and processes by cost.
+There is no GUI on purpose; it's a library.
 ## What it does
 - **Topology optimization** on a voxel grid (SIMP with a density filter,
@@ -60,23 +95,23 @@ licensed.
   clean-outs), plus milling and casting constraints.
 - **Field modelling**: a part as a graph of implicit fields, so skins,
   lattices (TPMS and strut) and fillets can be graded by stress or by
-  distance ([docs/fields.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/fields.md)).
+  distance ([docs/fields.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/fields.md)).
 - **Other analyses**: large deflection and buckling, compliant mechanisms,
   thermal loads, fatigue and creep estimates, lattice crush and impact,
   vibration, lattice-Boltzmann flow and conjugate heat for heat exchangers
   and cold plates, and layer-by-layer warp and residual stress
-  ([docs/BUILDSIM.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/BUILDSIM.md)).
+  ([docs/BUILDSIM.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/BUILDSIM.md)).
 - **CAD in and out**: a STEP or STL part becomes a draft study with its
-  bores and bolt patterns found ([docs/IMPORT.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/IMPORT.md)); results
+  bores and bolt patterns found ([docs/IMPORT.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/IMPORT.md)); results
   export as STL, 3MF, and STEP with exact planes and cylinders at the
   interfaces; a lattice part (strut cells, TPMS, graded, skinned or open)
-  goes to STEP lattice and all, as smooth patches on its true surface ([docs/fields.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/fields.md#the-lattice-itself-in-the-step-lattice-exact)).
+  goes to STEP lattice and all, as smooth patches on its true surface ([docs/fields.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/fields.md#the-lattice-itself-in-the-step-lattice-exact)).
 - **Mesh checks without a study** (`loadpath.checks`): wall thickness,
   overhangs, trapped voids, fluid-network isolation, component fit.
 ## How much to trust it
-Read [docs/VALIDATION.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/VALIDATION.md) before relying on a number.
+Read [docs/VALIDATION.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/VALIDATION.md) before relying on a number.
 In short:
 - **The solvers match analytic answers** in the test suite: beam
@@ -101,16 +136,16 @@ In short:
 ## Case studies
 Each states the problem, the setup, the numbers against a named baseline,
-and the limitations ([docs/case-studies/](https://github.com/connorwhite-online/loadpath/tree/v0.1.1/docs/case-studies/)).
+and the limitations ([docs/case-studies/](https://github.com/connorwhite-online/loadpath/tree/v0.1.3/docs/case-studies/)).
 | Case study | Result | Baseline |
 | --- | --- | --- |
-| [Eval suite](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/case-studies/evals.md) | 15 of 15 parts meet their requirements on a fresh analysis of the delivered mesh | Each part's own requirements |
-| [GRIP robot arm](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/case-studies/grip.md) | A load sweep over the arm's URDF found 3 of 5 links below SF 1; the regrown set meets SF 2 and is 18% smaller | GRIP's committed parts, made by an earlier loadpath with hand-written loads |
-| [Arm link cores](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/case-studies/arm-link-cores.md) | 42.2 g box section for FDM, 36.8 g hollow link for SLS, both at 0.5 mm and SF 3 | 54.9 g solid web; 54.2 g free-form topology optimization |
-| [Liquid cold plate](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/case-studies/cold-plate.md) | A die-graded gyroid ties streamline channels; channels filled with a graded gyroid run 7% cooler | Streamline channels and pin fins, equal metal and pressure drop |
-| [TPU finger](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/case-studies/tpu-finger.md) | up to 8 mm of tip curl for a 5 mm pull; a tendon finger grips 8x harder | A linear-model design (2.5 mm, then the solve fails); a plain bar of equal volume |
-| [Benchmarks](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/case-studies/benchmarks.md) | 1-5% less stiff than top88/top3d at equal material, in 57-60 iterations instead of 130-410 | Ports of top88 and top3d |
+| [Eval suite](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/case-studies/evals.md) | 15 of 15 parts meet their requirements on a fresh analysis of the delivered mesh | Each part's own requirements |
+| [GRIP robot arm](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/case-studies/grip.md) | A load sweep over the arm's URDF found 3 of 5 links below SF 1; the regrown set meets SF 2 and is 18% smaller | GRIP's committed parts, made by an earlier loadpath with hand-written loads |
+| [Arm link cores](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/case-studies/arm-link-cores.md) | 42.2 g box section for FDM, 36.8 g hollow link for SLS, both at 0.5 mm and SF 3 | 54.9 g solid web; 54.2 g free-form topology optimization |
+| [Liquid cold plate](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/case-studies/cold-plate.md) | A die-graded gyroid ties streamline channels; channels filled with a graded gyroid run 7% cooler | Streamline channels and pin fins, equal metal and pressure drop |
+| [TPU finger](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/case-studies/tpu-finger.md) | up to 8 mm of tip curl for a 5 mm pull; a tendon finger grips 8x harder | A linear-model design (2.5 mm, then the solve fails); a plain bar of equal volume |
+| [Benchmarks](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/case-studies/benchmarks.md) | 1-5% less stiff than top88/top3d at equal material, in 57-60 iterations instead of 130-410 | Ports of top88 and top3d |
 ## Command line and Python
@@ -129,7 +164,7 @@ numba), `gpu` (PyTorch for large grids on CUDA, Apple Silicon or CPU
 threads).
 Studies are validated against a published JSON Schema
-([`schema/`](https://github.com/connorwhite-online/loadpath/tree/v0.1.1/schema/)), and `loadpath check` returns every problem at once
+([`schema/`](https://github.com/connorwhite-online/loadpath/tree/v0.1.3/schema/)), and `loadpath check` returns every problem at once
 with its JSON path, so an agent can fix a study in one round trip. Long
 solves run as background jobs the agent polls. Each finished outcome has an
 `attention` list of numbers to look at before trusting it, and renders a
@@ -149,51 +184,51 @@ print(best.report["mass"], best.report["safetyFactor"])
 ## Documentation
-- [Quickstart](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/quickstart.md): install to a checked part.
-- [Tutorials](https://github.com/connorwhite-online/loadpath/tree/v0.1.1/docs/tutorials/): eleven end-to-end workflows, each run by CI
+- [Quickstart](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/quickstart.md): install to a checked part.
+- [Tutorials](https://github.com/connorwhite-online/loadpath/tree/v0.1.3/docs/tutorials/): eleven end-to-end workflows, each run by CI
   (an FDM bracket to 3MF, an SLS lattice with drain holes, a metal heat
   exchanger, an impact bumper, a STEP part round trip, LPBF distortion
   compensation, and more).
-- [Reference](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/loadpath/README.md): the method, every study field, the
+- [Reference](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/loadpath/README.md): the method, every study field, the
   report, the limits.
-- [Validation](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/VALIDATION.md): what each model has been checked
+- [Validation](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/VALIDATION.md): what each model has been checked
   against.
-- [Case studies](https://github.com/connorwhite-online/loadpath/tree/v0.1.1/docs/case-studies/) and [benchmarks](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/BENCHMARKS.md).
-- Topics: [fields](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/fields.md), [materials](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/materials.md),
-  [build simulation](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/BUILDSIM.md), [calibration](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/CALIBRATION.md),
-  [CAD import](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/IMPORT.md), [slicer](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/slicer.md),
-  [multi-material](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/MULTIMATERIAL.md), [impact](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/IMPACT.md),
-  [mechanisms](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/MECHANISMS.md).
+- [Case studies](https://github.com/connorwhite-online/loadpath/tree/v0.1.3/docs/case-studies/) and [benchmarks](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/BENCHMARKS.md).
+- Topics: [fields](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/fields.md), [materials](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/materials.md),
+  [build simulation](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/BUILDSIM.md), [calibration](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/CALIBRATION.md),
+  [CAD import](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/IMPORT.md), [slicer](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/slicer.md),
+  [multi-material](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/MULTIMATERIAL.md), [impact](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/IMPACT.md),
+  [mechanisms](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/MECHANISMS.md).
 <!-- gallery:start -->
 ## Examples
-The eval suite (15/15 passing): robot-arm links, brackets, a bell crank and a hook, each optimized and then checked by a fresh analysis of the delivered mesh. "Under envelope" compares with the design space filled solid, not with a conventionally designed part. These are simulated results; see [docs/VALIDATION.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/VALIDATION.md). Blue arrows are loads, red is keep-out. Every view, stress map and number is in [docs/GALLERY.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.1/docs/GALLERY.md).
+The eval suite (15/15 passing): robot-arm links, brackets, a bell crank and a hook, each optimized and then checked by a fresh analysis of the delivered mesh. "Under envelope" compares with the design space filled solid, not with a conventionally designed part. These are simulated results; see [docs/VALIDATION.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/VALIDATION.md). Blue arrows are loads, red is keep-out. Every view, stress map and number is in [docs/GALLERY.md](https://github.com/connorwhite-online/loadpath/blob/v0.1.3/docs/GALLERY.md).
 <table>
 <tr>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/arm-link.png" width="250"><br><b>arm-link</b><br><sub>55.3 g · 56% under envelope · SF 8.92 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/arm-link-cast.png" width="250"><br><b>arm-link-cast</b><br><sub>67.8 g · 46% under envelope · SF 20.11 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/arm-link-driven.png" width="250"><br><b>arm-link-driven</b><br><sub>48.4 g · 61% under envelope · SF 11.17 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/arm-link.png" width="250"><br><b>arm-link</b><br><sub>55.3 g · 56% under envelope · SF 8.92 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/arm-link-cast.png" width="250"><br><b>arm-link-cast</b><br><sub>67.8 g · 46% under envelope · SF 20.11 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/arm-link-driven.png" width="250"><br><b>arm-link-driven</b><br><sub>48.4 g · 61% under envelope · SF 11.17 · pass</sub></td>
 </tr>
 <tr>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/bell-crank.png" width="250"><br><b>bell-crank</b><br><sub>23.4 g · 84% under envelope · SF 3.22 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/grip-turret.png" width="250"><br><b>grip-turret</b><br><sub>140.4 g · 91% under envelope · SF 2.85 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/grip-upper-arm.png" width="250"><br><b>grip-upper-arm</b><br><sub>207.7 g · 88% under envelope · SF 7.33 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/bell-crank.png" width="250"><br><b>bell-crank</b><br><sub>23.4 g · 84% under envelope · SF 3.22 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/grip-turret.png" width="250"><br><b>grip-turret</b><br><sub>140.4 g · 91% under envelope · SF 2.85 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/grip-upper-arm.png" width="250"><br><b>grip-upper-arm</b><br><sub>207.7 g · 88% under envelope · SF 7.33 · pass</sub></td>
 </tr>
 <tr>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/grip-wrist-link-mjf.png" width="250"><br><b>grip-wrist-link-mjf</b><br><sub>40.2 g · 95% under envelope · SF 5.83 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/grip-wrist-yoke.png" width="250"><br><b>grip-wrist-yoke</b><br><sub>17.3 g · 93% under envelope · SF 3.88 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/mbb-beam.png" width="250"><br><b>mbb-beam</b><br><sub>13.8 g · 79% under envelope · SF 1.51 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/grip-wrist-link-mjf.png" width="250"><br><b>grip-wrist-link-mjf</b><br><sub>40.2 g · 95% under envelope · SF 5.83 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/grip-wrist-yoke.png" width="250"><br><b>grip-wrist-yoke</b><br><sub>17.3 g · 93% under envelope · SF 3.88 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/mbb-beam.png" width="250"><br><b>mbb-beam</b><br><sub>13.8 g · 79% under envelope · SF 1.51 · pass</sub></td>
 </tr>
 <tr>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/motor-mount.png" width="250"><br><b>motor-mount</b><br><sub>27.6 g · 69% under envelope · SF 3.65 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/printed-hook.png" width="250"><br><b>printed-hook</b><br><sub>24.2 g · 84% under envelope · SF 2.73 · pass</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/wall-bracket.png" width="250"><br><b>wall-bracket</b><br><sub>26.9 g · 94% under envelope · SF 2.85 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/motor-mount.png" width="250"><br><b>motor-mount</b><br><sub>27.6 g · 69% under envelope · SF 3.65 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/printed-hook.png" width="250"><br><b>printed-hook</b><br><sub>24.2 g · 84% under envelope · SF 2.73 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/wall-bracket.png" width="250"><br><b>wall-bracket</b><br><sub>26.9 g · 94% under envelope · SF 2.85 · pass</sub></td>
 </tr>
 <tr>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/cases/wall-bracket-fasteners.png" width="250"><br><b>wall-bracket-fasteners</b><br><sub>26.5 g · 94% under envelope · SF 2.57 · pass</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/cases/wall-bracket-fasteners.png" width="250"><br><b>wall-bracket-fasteners</b><br><sub>26.5 g · 94% under envelope · SF 2.57 · pass</sub></td>
 </tr>
 </table>
@@ -201,23 +236,23 @@ Where the load goes: von Mises stress, blue (none) to red (the part's 98th perce
 <table>
 <tr>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/stress/bell-crank.png" width="250"><br><sub>bell-crank</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/stress/grip-upper-arm.png" width="250"><br><sub>grip-upper-arm</sub></td>
-<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/stress/printed-hook.png" width="250"><br><sub>printed-hook</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/stress/bell-crank.png" width="250"><br><sub>bell-crank</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/stress/grip-upper-arm.png" width="250"><br><sub>grip-upper-arm</sub></td>
+<td align="center" valign="top"><img src="https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/stress/printed-hook.png" width="250"><br><sub>printed-hook</sub></td>
 </tr>
 </table>
 Two-fluid heat exchangers: two isolated channel networks through one sheet gyroid, with headers that print without support.
-![box-crossflow](https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/heatx/box-crossflow.png)
+![box-crossflow](https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/heatx/box-crossflow.png)
 Printed in metal, nothing inside can be supported: the channels end in peaked roofs instead of flat ceilings (red).
-![metal ceilings](https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/heatx/lpbf-ceilings.png)
+![metal ceilings](https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/heatx/lpbf-ceilings.png)
 Made another way: the same arm link printed, cast (no undercuts along z), 3-axis milled from both sides with a 6 mm cutter, and with members capped at 6 mm. Each meets SF 3 and 0.5 mm.
-![manufacturing constraints](https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.1/docs/images/manufacturing/arm-link.png)
+![manufacturing constraints](https://raw.githubusercontent.com/connorwhite-online/loadpath/v0.1.3/docs/images/manufacturing/arm-link.png)
 <sub>Rendered 2026-11-03 from `6c97b57` by `python docs/gallery.py`.</sub>

{loadpath-0.1.1 → loadpath-0.1.3}/loadpath/SKILL.md RENAMED Viewed

@@ -16,6 +16,18 @@ it, use the CLI below. Start from the nearest example (`loadpath_examples` /
 `loadpath examples`) rather than a blank page; `loadpath_schema` gives every
 field. `check` reports every problem at once: fix them all, then check again.
+## When to use it
+Use loadpath when the shape should come from the loads: brackets, links,
+mounts and housings around fixed interfaces, lattice-filled or graded parts,
+compliant mechanisms, heat exchangers and cold plates. Draw parts whose shape
+is set by function rather than load (enclosures, cosmetic covers, gears,
+threads) in CAD instead. loadpath doesn't replace CAD: model the interfaces
+there, let loadpath grow the material between them, and bring the result
+back as STEP or STL. Its edge over drawing the part yourself is that every
+result comes with an independent check (safety factor, deflection,
+buckling, printability) you can report to the user.
 ## Workflow
 1. **Model the interfaces exactly** (in build123d, CadQuery, or by hand):

loadpath 0.1.1__tar.gz → 0.1.3__tar.gz

loadpath 0.1.1tar.gz → 0.1.3tar.gz