v2026.7.19.1

PyAutoFit v2026.7.19.1

What's New

Breaking Changes

• fix: AggregateFITS leaked one file handle per HDU per result (crash at ~500 results) (#1386)
• fix: from_dict dropped dict entries with falsy values (0.0 parameters vanish on load) (#1384)
• feat: size-realistic test-mode bypass samples (PYAUTO_TEST_MODE_SAMPLES) (#1381)
  • Added-only and default-off: new env var PYAUTO_TEST_MODE_SAMPLES (default 4 — behaviour is byte-identical when unset) with the autoconf accessor re-exported as autofit.non_linear.test_mode.test_mode_samples. The private _build_fake_samples gains a vectorized N > 4 branch; the N == 4 literal branch is untouched. No existing public symbol changed.
• fix: sqlite database slicing + direct-write samples_summary (aggregator Phase D) (#1380)

New Features

• docs: pin Fitness's NaN guard contract — value-only, never gradient (#1392)
• refactor: public AbstractPaths.preserve_in_zip for post-completion artifacts (#1389) (#1390)

Bug Fixes

• fix: mark test-mode bypassed fits complete so they are resumable (#1387) (#1388)
• fix: EP projection — transform samples to base space + use log weights (#1383)

Internal

• chore: rename PyAutoConf → PyAutoNerves in docs/prose (#1396)
• refactor!: depend on and import autonerves (renamed from autoconf) (#1395)
• refactor: re-export autoconf config surface from autofit (#1394)
• Rename PyAutoBuild to PyAutoHands (#1393)
• perf: aggregator result-loading speedups (single-scan listing, cached summaries, faster samples parse) (#1376)
• feat(search): add batch_size to the multi-start gradient searches (#1374)

---

Full changelog: 2026.7.15.1...2026.7.19.1