uv dev environment, README writing pass, publish hygiene

- pyproject.toml + uv.lock: uv-managed dev environment (uv sync /
  uv run pytest). requirements.txt stays: KiCad builds the plugin's
  runtime venv from it and the PCM zip packages it.
- README: em-dash and run-on cleanup (34 -> 22, the rest deliberate),
  split the 120-word THT solder-joint sentence, fix the documented
  ADAPTIVE_MAX_CELL_UM value (2 mm -> 1 mm, config has 1000 um), fix
  the *Packaging / publishing* cross-reference, dev sections now use
  uv sync / uv run.
- LLM disclaimer: "most commits" carry the trailer (32 of 39), figures
  claim now excepts the hand-drawn hole cross-section, reference the
  UT3513+ measured-vs-computed validation.
- .gitignore: local AI-tooling artifacts; deploy scripts exclude
  pyproject.toml/uv.lock; error-figure title punctuation aligned with
  the other window titles.

.gitignore

@@ -3,3 +3,11 @@ __pycache__/
 *.pyc
 .pytest_cache/
 dist/
+
+# local AI-tooling artifacts, never publish
+.claude/
+.claude-flow/
+.swarm/
+.mcp.json
+CLAUDE.md
+ruvector.db

README.md

@@ -6,8 +6,8 @@ between two contacts, **single- or multi-layer**: the chosen net's fills
 solved as coupled finite-difference sheets linked by the net's **via
 and through-hole-pad barrels** (18 µm plating, configurable). At a user-set **frequency** the exact 1D foil/barrel
 skin-effect correction is applied (AC results are a rigorous lower
-bound — see *Model & limits*). Shows per-layer rasterized maps,
+bound; see *Model & limits*). Shows per-layer rasterized maps,
-potential, current density, and **power density**, reports **per-via
+potential, current density, and **power density**, and reports **per-via
 currents** (via ampacity!) and total dissipation at a **selectable test
 current**. PNGs + a text summary are saved per run.
 
@@ -15,7 +15,7 @@ current**. PNGs + a text summary are saved per run.
 *Real output on a synthetic two-layer net: current from a soldered
 THT-pad contact (V+, injected at the drill-wall ring) squeezes past a
 notch in the F.Cu pour, transfers through the stitching-via field into
-the B.Cu pour and leaves at the V− lug — per-via currents and the
+the B.Cu pour and leaves at the V− lug. Per-via currents and the
 hottest via are reported.*
 
 ![Potential on the two-layer demo net](docs/img/demo-potential.png)
@@ -35,7 +35,7 @@ SWIG API. Requires KiCad **10.0.1+**.
    on Windows or `/usr/bin/python3` on Linux (after a 9→10 upgrade it
    can point at KiCad 9).
 3. **Deploy** (dev checkout; end users install the PCM zip instead, see
-   *Packaging*):
+   *Packaging / publishing*):
    ```powershell
    powershell -ExecutionPolicy Bypass -File deploy.ps1        # junction (dev)
    powershell -ExecutionPolicy Bypass -File deploy.ps1 -Mode Copy
@@ -52,13 +52,13 @@ SWIG API. Requires KiCad **10.0.1+**.
 ## Usage
 
 1. Mark the current-injection terminals. Each terminal may have
-   **multiple parts** (all merged into one externally-bonded contact):
+   **multiple parts** (all merged into one externally bonded contact):
    - **V+ rectangles on `User.1`**, **V− rectangles on `User.2`**
      (marker layers, configurable via `ELECTRODE_POS_LAYER` /
      `ELECTRODE_NEG_LAYER`), any number per side, axis-aligned;
    - **pads and vias** (SMD pad: real copper shape on its own layer;
-     through-hole pads and vias become **barrel contacts** — the current
+     through-hole pads and vias become **barrel contacts**: the current
-     enters at the drill wall on every spanned layer, see below) —
+     enters at the drill wall on every spanned layer, see below);
      selected pads/vias fill a side that has no rectangles;
    - legacy: exactly 2 selected contacts with no marker rectangles still
      works; empty selection scans the whole board's marker layers.
@@ -93,7 +93,7 @@ SWIG API. Requires KiCad **10.0.1+**.
   dialog's **"capped up to drill"** threshold (default
   `CAP_MAX_DRILL_MM = 0.5`) keep open mouths even with capping
   selected. Layer-to-layer the cap never matters at DC (it is in
-  parallel with the annular-ring contact, not in series) — the checkbox
+  parallel with the annular-ring contact, not in series), so the checkbox
   only affects in-plane conduction across outer-layer mouths. Sub-cell
   mouths scale their cells' sheet conductance by the true covered
   fraction (4×4 supersampling), so coarse grids see the correct small
@@ -103,18 +103,18 @@ SWIG API. Requires KiCad **10.0.1+**.
   oblong pads, fetched from KiCad; the outer shape stands in for inner
   rings) are stamped onto every included layer, and every **populated**
   pad carries its full **soldered joint** on its SOLDER side (opposite
-  the component; the component-side pad face stays bare): the hole
+  the component; the component-side pad face stays bare). The hole
   holds the **component lead** (a cylinder of drill −
   `THT_LEAD_CLEARANCE_MM`, resistivity `THT_LEAD_RHO_OHM_M`, copper by
-  default — raise it for brass/steel leads) **plus solder** in the
+  default; raise it for brass/steel leads) **plus solder** in the
-  remaining annulus, both in parallel with the plating; the mouth
+  remaining annulus, both in parallel with the plating. The mouth
-  copper stays conducting (it stands in for the plug — conservative,
+  copper stays conducting: it stands in for the solder plug, which is
-  the plug is worth far more than the foil); the pad face gets the
+  worth far more than the foil, so this is conservative. The pad face
-  average-thickness solder coat (exact pad shape) and the
+  gets the average-thickness solder coat (exact pad shape) and the
   protruding-lead cone (see barrel contacts below; on oblong pads the
   cone tapers within the inscribed circle). Whether a hole is a via or
   a THT pad, the owning footprint's side, and its **Do not populate**
-  flag are all read from KiCad — **DNP pads** get an **open hole** and
+  flag are all read from KiCad. **DNP pads** get an **open hole** and
   a plating-only barrel, no joint. At f > 0 the thickness scaling is
   applied multiplicatively to the skin-corrected sheet conductance
   (approximation). Per layer a barrel attaches to
@@ -194,7 +194,7 @@ SWIG API. Requires KiCad **10.0.1+**.
   (`SKIN_SIDES = 1` in config: plane facing a return plane; `2` =
   isolated foil), and the analogous correction for the 18 µm barrel wall.
   Enter one frequency per run (e.g. a switching harmonic, with its RMS
-  amplitude as the test current) — suffixes `k`/`M` accepted.
+  amplitude as the test current); suffixes `k`/`M` are accepted.
   **Caveat:** only through-thickness crowding is modeled. Lateral
   (proximity-effect) redistribution needs a magneto-quasistatic solver
   and is not captured — since the resistance-driven distribution is the
@@ -206,15 +206,15 @@ SWIG API. Requires KiCad **10.0.1+**.
   not the geometric foil thickness.
 - 5-point FDM per layer on an auto-sized shared grid (~2 M fine cells
   with the uniform grid; ~8 M with the adaptive grid, whose unknown
-  count no longer scales with them). Direct sparse solve up to 500 k
+  count no longer scales with the fine-cell count). Direct sparse solve
-  unknowns, AMG-preconditioned CG (pyamg) above — Jacobi-CG if pyamg is
+  up to 500 k unknowns, AMG-preconditioned CG (pyamg) above (Jacobi-CG
-  missing. Discretization error typically ≲ 2 % at defaults — halve the
+  if pyamg is missing). Discretization error typically ≲ 2 % at
-  cell size and compare to judge convergence.
+  defaults; halve the cell size and compare to judge convergence.
 - **Adaptive cells** (dialog checkbox, **on by default**;
   `ADAPTIVE_CELLS`):
   solves on a 2:1-balanced quadtree — fine cells at copper boundaries,
   electrodes, traces, via mouths and buildup, blocks up to
-  `ADAPTIVE_MAX_CELL_UM` (2 mm) in plane interiors (`ADAPTIVE_GUARD`
+  `ADAPTIVE_MAX_CELL_UM` (1 mm) in plane interiors (`ADAPTIVE_GUARD`
   sets the clearance a block needs to grow). The **minimum element size
   is the grid cell size itself** (auto / dialog / `CELL_UM_OVERRIDE`);
   the uniform limit reproduces the normal grid exactly. Large
@@ -246,20 +246,20 @@ accordingly more trustworthy than absolute numbers.
 Every run writes `geometry_dump.json`; re-solve without KiCad:
 
 ```powershell
-.venv\Scripts\python.exe -m fill_resistance.standalone dump.json `
+uv run python -m fill_resistance.standalone dump.json `
     [--current 40] [--cell-um 50] [--layers F.Cu,In1.Cu] [--no-show] `
     [--out DIR] [--force-iterative]
 ```
 
-Dev environment, tests, headless extraction (Windows shown; on
+Dev environment, tests, headless extraction — [uv](https://docs.astral.sh/uv/)
-Linux/macOS use `.venv/bin/python`):
+manages the venv from `pyproject.toml`/`uv.lock` (`requirements.txt`
+stays: KiCad builds the plugin's runtime venv from it):
 
 ```powershell
-uv venv --python 3.11 .venv
+uv sync                                              # one-time env setup
-uv pip install --python .venv\Scripts\python.exe kicad-python numpy scipy pyamg matplotlib pytest
+uv run pytest -q                                     # incl. exact analytic cases
-.venv\Scripts\python.exe -m pytest tests -q          # incl. exact analytic cases
+uv run python tools/api_probe.py                     # IPC API probe vs live KiCad
-.venv\Scripts\python.exe tools\api_probe.py          # IPC API probe vs live KiCad
+uv run python -m fill_resistance.board_io dump.json [NET]  # extract only
-.venv\Scripts\python.exe -m fill_resistance.board_io dump.json [NET]  # extract only
 ```
 
 ## Packaging / publishing
@@ -272,7 +272,7 @@ filled in. To publish: upload the zip to a release, set `download_url`
 registry copy as `packages/th.co.b4l.fill-resistance/metadata.json` in a
 merge request to <https://gitlab.com/kicad/addons/metadata>. Icons are
 regenerated with `python tools/gen_icons.py`; the README figures in
-`docs/img/` with `.venv\Scripts\python.exe tools\gen_readme_figs.py`
+`docs/img/` with `uv run python tools/gen_readme_figs.py`
 (real solver output on small synthetic boards, plus the hand-drawn
 hole cross-section).
 
@@ -294,18 +294,20 @@ GPL-3.0-or-later — see [LICENSE](LICENSE).
 
 ## LLM disclaimer
 
-This plugin was developed with an LLM — Anthropic's **Claude** (Claude
+This plugin was developed with an LLM: Anthropic's **Claude** (Claude
 Code, model Claude Fable 5). The physics model, solver, tests, tooling
-and this documentation (including the figures, which are generated by
+and this documentation (including the figures; all but the hand-drawn
-the solver itself) were written by the model, feature by feature, under
+hole cross-section are generated by the solver itself) were written by
-human direction and review (janik / B4L); commits carry a
+the model, feature by feature, under human direction and review
-`Co-Authored-By: Claude` trailer.
+(janik / B4L); most commits carry a `Co-Authored-By: Claude` trailer.
 
 What keeps this honest: the test suite pins the numerics to exact
 analytic references (strip and annulus resistances, the acosh spreading
 resistance of two circular contacts, skin-effect limits, power-balance
-identities) and to convergence/regression checks — run it with
+identities) and to convergence/regression checks; run it with
-`pytest tests`. Nevertheless, an LLM wrote this: read *Model & limits*
+`uv run pytest`. Real boards were measured against a UT3513+ micro-ohm
+meter (see *Measured vs. computed* above). Nevertheless, an LLM wrote
+this: read *Model & limits*
 critically, treat surprising numbers with the usual engineering
 suspicion, and cross-check against a hand estimate before trusting a
 result with hardware. Bug reports are very welcome.

deploy.ps1

@@ -30,7 +30,8 @@ if ($Mode -eq 'Junction') {
     New-Item -ItemType Junction -Path $dst -Target $src | Out-Null
     Write-Host "junction created: $dst -> $src"
 } else {
-    $exclude = @('.venv', '.git', 'tests', 'tools', '__pycache__', '.pytest_cache')
+    $exclude = @('.venv', '.git', 'tests', 'tools', '__pycache__', '.pytest_cache',
+                 'pyproject.toml', 'uv.lock')
     New-Item -ItemType Directory -Force $dst | Out-Null
     Get-ChildItem $src -Force | Where-Object { $exclude -notcontains $_.Name } |
         ForEach-Object { Copy-Item $_.FullName -Destination $dst -Recurse -Force }

fill_resistance/plots.py

@@ -416,7 +416,7 @@ def fig_power(result, stack, e1, e2, problem):
 def fig_error(message: str):
     fig, ax = plt.subplots(figsize=(9, 4.5), layout="constrained")
     ax.axis("off")
-    ax.set_title("Fill Resistance — ERROR", color="#b02a2a",
+    ax.set_title("Fill Resistance - ERROR", color="#b02a2a",
                  fontsize=14, fontweight="bold", loc="left")
     wrapped = "\n".join(
         textwrap.fill(line, width=90) for line in message.splitlines()

pyproject.toml

@@ -0,0 +1,28 @@
+# Development environment only (uv sync / uv run). The KiCad plugin
+# manager builds the runtime venv itself from requirements.txt — keep
+# the dependency list there in sync with [project.dependencies].
+[project]
+name = "fill-resistance"
+version = "1.1.0"
+description = "DC/AC resistance of copper zone fills and traces between two contacts (KiCad 10 plugin)"
+license = "GPL-3.0-or-later"
+requires-python = ">=3.11"
+dependencies = [
+    "kicad-python>=0.7.0",
+    "numpy",
+    "scipy",
+    "pyamg",
+    "matplotlib",
+    "PySide6",
+]
+
+[dependency-groups]
+dev = [
+    "pytest",
+]
+
+[tool.uv]
+package = false
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]

tools/deploy.py

@@ -18,7 +18,7 @@ from pathlib import Path
 ROOT = Path(__file__).resolve().parent.parent
 COPY_EXCLUDE = {".venv", ".git", "tests", "tools", "dist", "resources",
                 "__pycache__", ".pytest_cache", "conftest.py", "deploy.ps1",
-                ".gitignore", "metadata.json"}
+                ".gitignore", "metadata.json", "pyproject.toml", "uv.lock"}
 
 
 def plugins_dir(kicad_version: str) -> Path: