[roytsmart]

Expands the README and the documentation introduction now that least_squares() has merged (#4), following the layout of ndfilters.

Changes

Both the README and docs/index.rst now have:

• A runnable quickstart. The old README example had ... placeholders. The new one fits a Gaussian to a thousand noisy lines and evaluates the best-fit models on a finer grid with ndoptimize.evaluate(). Each fit starts at the peak of its own line, which also shows a per-fit initial guess. A shared guess at the center let 2 of the 1,000 lines, the ones 3σ off center, run away. The documentation plots three of the lines with their fits, and the README shows that figure from Read the Docs.
• Writing a model. An exponential decay, written as a Numba model with its Jacobian and compiled with @ndoptimize.model, plus a note on cache=False for models defined in the interactive interpreter. The documentation fits it to a hundred decays with bounds and plots the fitted scales against the truth, with error bars.
• Differences from scipy.optimize.least_squares(), like the scipy.ndimage section of ndfilters. The two find the same minimum for the same objective, as the tests check, but differ in:
  • the Jacobian, which a model computes itself;
  • noise and priors, which are arguments here, along with the covariance in the result;
  • bounds, which use projection here and trust regions in SciPy;
  • the convergence test, the decrease in chi-squared that a Gauss-Newton step predicts here;
  • what isn't supported: robust losses and sparse Jacobians.
• A "Custom models" feature, for models sharing one cached kernel.
• A citation section pointing to CITATION.cff, with a BibTeX entry, in the pre-DOI form of colorsynth#24.

The documentation also lists what the package provides: least_squares, evaluate, model/Model and models.

Review fixes (8b65ce6)

• The broadcasting feature says each fit can have its own independent variable, and that the noise correlation is shared by every fit.
• The SciPy differences also cover:
  • model arguments: extra data goes in more columns of x, not args/kwargs;
  • chi2 versus SciPy's cost;
  • max_iterations and iterations counting accepted steps, not evaluations.
• The convergence quantity is now called the decrease in chi-squared that a Gauss-Newton step predicts, which is half the square of the Newton decrement. The tolerance docstring of least_squares() now uses the same wording, and the max_iterations and iterations docstrings say they count accepted steps. This is the PR's only change to the package.
• The cache=False note now comes before the model example, which fails when pasted into the interactive interpreter.
• The quickstart's noise is called noise rather than sigma, since sigma is the width of the Gaussian model, and a comment gives the shape of x.
• The README says its figure comes from the documentation, and an index.rst comment keeps that cell first, since the image's name depends on cell order.

Checks

• The docs build locally without warnings. Every cross-reference resolves, as do the README links to API pages.
• Both README code blocks run. All 1,000 quickstart fits converge in at most 12 steps, and the center pulls have a standard deviation of 0.99.

Follow-ups

• The README figure is served from Read the Docs latest (_images/index_0_0.png), so it shows only after this merges and latest rebuilds.
• After the first release is archived on Zenodo, a follow-up PR adds the concept DOI to CITATION.cff, the README badge and the citation sections, as for the other sun-data packages.

🤖 Generated with Claude Code

https://claude.ai/code/session_01PZRXiFhGzzrF87vsFg73B9

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.61%. Comparing base (61296fd) to head (8b65ce6).

Additional details and impacted files

@@           Coverage Diff           @@
##             main       #5   +/-   ##
=======================================
  Coverage   99.61%   99.61%           
=======================================
  Files           8        8           
  Lines         782      782           
=======================================
  Hits          779      779           
  Misses          3        3           

Flag	Coverage Δ
unittests	99.61% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.