docs: reflect normalization fixes, detector cleanup and CLI overrides

- README: remove stale detector_response.py tree entry and the
  detector-models highlight (module removed); document per-branch
  measured transition endpoints from paceENSDF and JSON-flow CLI
  override semantics
- TODO: drop the DEFAULT_JSON_SCHEMA detector-keys clause (removed)
- data/custom_input_example.json: strip dead detector_* keys

README.md

@@ -60,8 +60,7 @@ beta_shape_pnpi/
 │       │   ├── finite_size.py           ✓ L0 expansion + charge distribution U term
 │       │   ├── screening.py             ✓ Atomic electron screening (ratio method)
 │       │   ├── exchange.py              ✓ Hayen-2018 empirical fit coefficients
-│       │   ├── radiative.py             ✓ Outer radiative corrections, soft-photon resummation
-│       │   └── detector_response.py     ✓ Analytical detector response models
+│       │   └── radiative.py             ✓ Outer radiative corrections, soft-photon resummation
 ├── beta_spectrum/output/    ← output, plotting, visualization, CSV export
 │   ├── __init__.py          # BetaSpectrumAnalyzer, CSVEmitter, Visualizer
 │   ├── core.py              # BetaSpectrumAnalyzer base class
@@ -191,9 +190,9 @@ ______________________________________________________________________
 For a complete list of implemented and planned features, see [TODO.md](TODO.md). Key highlights:
 
 - **Physics corrections**: phase space, Fermi function, finite size, screening, exchange, radiative (with delta_cut resummation)
-- **Detector response**: Gaussian, Gaussian+tail, Tikhonov models with tabulated support
+- **Detector response**: not implemented — `SpectrumModel` keeps a duck-typed optional convolution hook; the analytical models were removed (restore-or-delete decision tracked in TODO.md)
 - **Multi-branch decay**: intensity-weighted sum of branch spectra, per-branch calculators with branch-specific endpoint, transition type (from ENSDF), and `--intensity-cutoff` CLI filter
 - **Analysis tools**: χ² fitter (SpectrumFitter + CurveFitter), C(W) extraction pipeline, Kurie plot analysis, g_V/g_A extraction
 - **Experimental data**: ExpSpectrum container, energy calibration, peak fitting, dead-time/pile-up correction
-- **CLI**: `bs_pnpi --nuclide Tc99` — full paceENSDF integration (auto-lookup of Z, A, Q-value, transition type, half-life, branches, forbiddenness from ENSDF)
-- **paceENSDF integration**: `get_decay_info_from_paceENSDF()`, `create_config_from_source("paceENSDF", nuclide="Tc99")`, `_parse_nuclide_symbol()` for Z=1..98, FORBIDDENNESS_MAP, `--nuclide` CLI flag
+- **CLI**: `bs_pnpi --nuclide Tc99` — full paceENSDF integration (auto-lookup of Z, A, Q-value, transition type, half-life, branches, forbiddenness from ENSDF); for JSON input the file's values are used and CLI flags (`--e-step`, `--neutrino-mass`, `--intensity-cutoff`, `--csv-separator`, `--normalize-to-one`) override them only when explicitly given
+- **paceENSDF integration**: `get_decay_info_from_paceENSDF()`, `create_config_from_source("paceENSDF", nuclide="Tc99")`, `_parse_nuclide_symbol()` for Z=1..98, FORBIDDENNESS_MAP, `--nuclide` CLI flag; per-branch endpoints are the measured transition endpoint energies from ENSDF (max endpoint across branches defines the spectrum endpoint)

TODO.md

@@ -32,7 +32,7 @@ beta_spectrum/              ← theory-only core
 │   ├── screening.py
 │   ├── exchange.py
 │   ├── radiative.py
-│   └── detector_response.py   ← stays here (part of theory)
+│   └── (detector_response.py — removed; restore-or-delete pending below)
 └── nuclear_data.py         ← paceENSDF integration
 
 exp_data/                   ← experimental data handling
@@ -102,9 +102,10 @@ output/                     ← output, visualization, CSV export
 - [x] README updated with new architecture diagram and Quick Start examples
 - [x] Version bumped to 0.4.0
 - [ ] Decide: restore or delete the `detector_response` module — it was removed from
-  `beta_spectrum/components/` together with its tests, but `fitter/model.py` keeps a
-  duck-typed optional hook and `DEFAULT_JSON_SCHEMA` keeps the detector keys;
-  restore from git history (commits up to `5f99566`) if convolution is needed again
+  `beta_spectrum/components/` together with its tests; `fitter/model.py` keeps a
+  duck-typed optional hook only (`DEFAULT_JSON_SCHEMA` detector keys and the CLI
+  dry-run detector line were removed with audit cleanup).
+  Restore from git history (commits up to `5f99566`) if convolution is needed again
 
 ______________________________________________________________________
 

data/custom_input_example.json

@@ -5,14 +5,5 @@
     "A_number": 99,
     "endpoint_MeV": 0.294,
     "transition_type": "A",
-    "e_step_MeV": 0.001,
-    "use_detector_response": false,
-    "detector_model": "gaussian",
-    "detector_sigma_a_keV": 1.0,
-    "detector_sigma_b": 0.0,
-    "detector_tail_fraction": 0.0,
-    "detector_tau_keV": 5.0,
-    "detector_fano_factor": 0.12,
-    "detector_n_channels": 4096,
-    "detector_channel_energy_range": [0.0, 0.35]
+    "e_step_MeV": 0.001
 }