nanodsp 0.1.3__tar.gz

nanodsp-0.1.3/.gitignore

@@ -0,0 +1,214 @@
+.DS_Store
+
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[codz]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py.cover
+.hypothesis/
+cover/
+
+# general cache
+.cache
+.*_cache
+.*-cache
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+#poetry.toml
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#   pdm recommends including project-wide configuration in pdm.toml, but excluding .pdm-python.
+#   https://pdm-project.org/en/latest/usage/project/#working-with-version-control
+#pdm.lock
+#pdm.toml
+.pdm-python
+.pdm-build/
+
+# pixi
+#   Similar to Pipfile.lock, it is generally recommended to include pixi.lock in version control.
+#pixi.lock
+#   Pixi creates a virtual environment in the .pixi directory, just like venv module creates one
+#   in the .venv directory. It is recommended not to include this directory in version control.
+.pixi
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.envrc
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+# Abstra
+# Abstra is an AI-powered process automation framework.
+# Ignore directories containing user credentials, local state, and settings.
+# Learn more at https://abstra.io/docs
+.abstra/
+
+# Visual Studio Code
+#  Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore 
+#  that can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore
+#  and can be added to the global gitignore or merged into this file. However, if you prefer, 
+#  you could uncomment the following to ignore the entire vscode folder
+# .vscode/
+
+# PyPI configuration file
+.pypirc
+
+# Cursor
+#  Cursor is an AI-powered code editor. `.cursorignore` specifies files/directories to
+#  exclude from AI features like autocomplete and code analysis. Recommended for sensitive data
+#  refer to https://docs.cursor.com/context/ignore-files
+.cursorignore
+.cursorindexingignore
+
+# Marimo
+marimo/_static/
+marimo/_lsp/
+__marimo__/
+
+# agents
+.claude
+CLAUDE.md
+AGENTS.md
+GEMINI.md
+

nanodsp-0.1.3/CHANGELOG.md

@@ -0,0 +1,295 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [Unreleased]
+
+## [0.1.3]
+
+### Added
+
+- **Virtual Analog filter bindings** (`nanodsp._core.vafilters`, `nanodsp.effects`) -- 6 Faust-derived analog-modeled filters
+  - `MoogLadder` -- 24 dB/oct Moog ladder lowpass with resonance
+  - `MoogHalfLadder` -- 12 dB/oct Moog half-ladder lowpass
+  - `DiodeLadder` -- 24 dB/oct diode ladder lowpass with internal soft clipping
+  - `Korg35LPF` -- 24 dB/oct Korg-35 lowpass
+  - `Korg35HPF` -- 24 dB/oct Korg-35 highpass
+  - `OberheimSVF` -- multi-mode state-variable filter with 4 simultaneous outputs (LPF, HPF, BPF, BSF)
+  - Python wrappers: `va_moog_ladder`, `va_moog_half_ladder`, `va_diode_ladder`, `va_korg35_lpf`, `va_korg35_hpf`, `va_oberheim`
+  - 44 tests (`tests/test_vafilters.py`)
+
+- **Band-limited oscillator bindings** (`nanodsp._core.bloscillators`, `nanodsp.synthesis`) -- 5 anti-aliased oscillator algorithms
+  - `PolyBLEP` -- polynomial band-limited step oscillator with 14 waveforms (sine, cosine, triangle, square, rectangle, sawtooth, ramp, modified triangle/square, half/full-wave rectified sine, triangular pulse, trapezoid fixed/variable)
+  - `BlitSaw` -- BLIT (band-limited impulse train) sawtooth with configurable harmonics
+  - `BlitSquare` -- BLIT square wave with DC blocker
+  - `DPWSaw` -- DPW (differentiated parabolic wave) sawtooth
+  - `DPWPulse` -- DPW pulse with variable duty cycle
+  - Python wrappers: `polyblep`, `blit_saw`, `blit_square`, `dpw_saw`, `dpw_pulse`
+  - 83 tests (`tests/test_bloscillators.py`)
+
+- **FX DSP algorithms** (`nanodsp._core.fxdsp`, `nanodsp.effects`, `nanodsp.synthesis`) -- 6 algorithms from cleaned/rewritten third-party sources
+  - `HardClipper` -- first-order antiderivative antialiased hard clipping
+  - `SoftClipper` -- first-order antiderivative antialiased soft clipping (sin saturation)
+  - `Wavefolder` -- second-order antiderivative antialiased wavefolding
+  - `SchroederReverb` -- classic 4 parallel feedback combs + 2 series allpasses with optional LFO modulation
+  - `MoorerReverb` -- Schroeder extension with 18-tap early reflections delay network
+  - `MinBLEP` -- minimum band-limited step oscillator (saw, reverse saw, square, triangle) with precomputed 2048-element table at 64x oversampling
+  - `PsolaShifter` -- PSOLA pitch shifting with autocorrelation pitch detection and grain-based resynthesis
+  - `FormantFilter` -- 3 cascaded bandpass biquads tuned to vowel formant frequencies (A/E/I/O/U) with blending
+  - Python wrappers: `aa_hard_clip`, `aa_soft_clip`, `aa_wavefold`, `schroeder_reverb`, `moorer_reverb`, `formant_filter`, `psola_pitch_shift`, `minblep`
+  - 69 tests (`tests/test_fxdsp.py`)
+
+- **Multi-order IIR filter design** (`nanodsp._core.iirdesign`, `nanodsp.effects`) -- 5 classical filter families via DspFilters (Vinnie Falco, MIT)
+  - Butterworth (maximally flat passband)
+  - Chebyshev Type I (passband ripple, sharper rolloff)
+  - Chebyshev Type II (stopband ripple, flat passband)
+  - Elliptic (sharpest transition, ripple in both bands)
+  - Bessel (linear phase, minimal ringing)
+  - Each family supports lowpass, highpass, bandpass, bandstop configurations
+  - Orders 1-16, returning SOS (second-order sections) coefficients
+  - `IIRFilter` class for stateful processing with `setup()`/`process()`/`reset()`/`sos()`
+  - `iir_design()` returns SOS coefficient array `[n_sections, 6]`
+  - `iir_filter()` applies multi-order IIR filter to AudioBuffer
+  - 41 tests (`tests/test_iirdesign.py`)
+
+- **Pure NumPy DSP algorithms** -- 7 new functions for API completeness without scipy dependency
+  - `ops.xcorr(buf_a, buf_b=None)` -- FFT-based cross-correlation (or autocorrelation in single-arg form)
+  - `ops.hilbert(buf)` -- amplitude envelope via analytic signal (FFT method)
+  - `ops.envelope(buf)` -- alias for `hilbert`
+  - `ops.median_filter(buf, kernel_size=3)` -- per-channel median filtering via stride tricks
+  - `ops.lms_filter(buf, ref, filter_len=32, step_size=0.01, normalized=True)` -- NLMS adaptive filter returning `(output, error)`
+  - `effects.agc(buf, target_level, max_gain_db, average_len, attack, release)` -- automatic gain control with asymmetric attack/release
+  - `analysis.gcc_phat(buf, ref, sample_rate)` -- GCC-PHAT time-delay estimation returning `(delay_seconds, correlation)`
+
+- **GrainflowLib bindings** (`nanodsp._core.grainflow`) -- granular synthesis engine (header-only, MIT license)
+  - `GfBuffer` -- buffer wrapper bridging numpy `[channels, frames]` arrays to GrainflowLib's internal AudioFile storage
+  - `GrainCollection` -- core multi-grain granulator with block-based processing, parameter control (enum and string reflection), buffer assignment, stream management, and auto-overlap
+  - `Panner` -- stereo grain panning with three modes (bipolar, unipolar, stereo) using equal-power quarter-sine interpolation
+  - `Recorder` -- live recording into buffers with overdub, freeze, sync, and multi-band filter support
+  - `Phasor` -- clock generator for grain triggering (continuous-phase ramp [0, 1))
+  - 37 enum constants: `PARAM_*` (23 parameter names), `PTYPE_*` (5 parameter types), `STREAM_*` (4 stream modes), `BUF_*` (6 buffer types), `BUFMODE_*` (3 buffer modes), `PAN_*` (3 pan modes)
+  - String-based parameter reflection (e.g. `"delayRandom"`, `"rateOffset"`, `"channelMode"`)
+  - All processing methods release the GIL for thread safety
+  - 49 tests (`tests/test_grainflow.py`)
+  - Patched two GrainflowLib upstream bugs for `SigType=float`: `gf_utils::mod` template deduction failure, `stream` method vs member access
+
+- **Demo scripts** (`demos/`) -- 16 runnable demo scripts showcasing the full API surface
+  - `demo_filters.py` -- 13 biquad filter variants (lowpass, highpass, bandpass, notch, peak, shelving)
+  - `demo_modulation.py` -- 10 modulation effects (chorus, flanger, phaser, tremolo)
+  - `demo_distortion.py` -- 14 distortion/saturation effects (overdrive, wavefold, bitcrush, decimator, saturate, fold)
+  - `demo_reverb.py` -- 12 reverb algorithms (FDN presets, ReverbSc, STK freeverb/jcrev/nrev/prcrev)
+  - `demo_dynamics.py` -- 9 dynamics processors (compression, limiting, gating, parallel/multiband compression)
+  - `demo_delay.py` -- 8 delay effects (stereo delay, ping-pong, slapback, echo)
+  - `demo_pitch.py` -- 10 pitch shifters (time-domain and spectral at various intervals)
+  - `demo_spectral.py` -- 12 spectral transforms (time stretch, phase lock, spectral gate, tilt EQ, freeze)
+  - `demo_daisysp_filters.py` -- 21 DaisySP filter variants (SVF, ladder, moog, tone, modal, comb)
+  - `demo_composed.py` -- 13 composed effects (autowah, sample rate reduce, DC block, exciter, de-esser, vocal chain, mastering, STK chorus)
+  - `demo_spectral_extra.py` -- 8 additional spectral transforms (denoise, EQ match, spectral morph)
+  - `demo_ops.py` -- 29 core DSP operations (delay, vibrato, convolution, envelopes, fades, panning, stereo widening, crossfade, normalization, trim, oversample)
+  - `demo_resample.py` -- 6 resampling variants (madronalib and FFT methods at 22k/48k/96k)
+  - `demo_synthesis.py` -- 44 synthesis sounds (oscillators, FM, formant, noise, drums, physical modeling, STK instruments, sequence) -- no input file required
+  - `demo_analysis.py` -- audio analysis printout (loudness, spectral features, pitch detection, onset detection, chromagram) -- no audio output
+  - `demo_grainflow.py` -- 7 granular synthesis variants (basic cloud, dense cloud, pitch shift up/down, sparse stochastic, stereo panned, recorder)
+  - `demo_fxdsp.py` -- 28 FX DSP outputs: antialiased waveshaping (6), Schroeder/Moorer reverbs (6), formant vowels (5), PSOLA pitch shifts (6), minBLEP waveforms (5)
+  - `demo_iir_filters.py` -- 23 IIR filter outputs: Butterworth (6), Chebyshev I (4), Chebyshev II (3), Elliptic (3), Bessel (4), order comparison (3)
+  - All file-processing scripts accept positional `infile`, optional `-o`/`--out-dir` (default `build/demo-output/`), and `-n`/`--no-normalize` to skip peak normalization
+  - Peak normalization (0 dBFS) applied by default to prevent clipping on PCM output
+- `make demos` target -- runs all 18 demo scripts in sequence (`DEMO_INPUT=demos/s01.wav` by default)
+
+### Fixed
+
+- **Moorer reverb early reflections routing** -- early reflections now bypass comb filters and mix directly to output (classic Moorer design)
+- **Moorer reverb delay read direction** -- fixed EarlyReflections reading forward (unwritten buffer) instead of backward (past samples)
+- **Schroeder reverb bugs** (from original source) -- all 4 combs incorrectly used same filter instance; allpass path used uninitialized variable
+- **DPW oscillator startup transient** -- seeded differentiator state in `reset()` to eliminate first-sample amplitude spike (~25x) caused by uninitialized `last_value_`
+- **Faust VA filter NaN** -- seeded parameter smoothing registers in MoogLadder, MoogHalfLadder, and DiodeLadder to prevent `log10(~0)` producing -inf/NaN on first samples
+
+## [0.1.2]
+
+### Changed
+
+- **GIL release in C++ bindings** -- all ~160 processing functions across 6 binding files now release the Python GIL during computation via `nb::gil_scoped_release`, enabling true multi-threaded parallelism
+  - `_core_signalsmith.cpp` -- Biquad, FFT, RealFFT, Delay, LFO, envelope, STFT, Oversampler processing
+  - `_core_daisysp.cpp` -- 73 functions: oscillators, filters, effects, dynamics, control, noise, drums, physical modeling, utility
+  - `_core_stk.cpp` -- generators, filters (via macro), reverbs (via macro), instruments (via macro), effects, Guitar, Twang
+  - `_core_madronalib.cpp` -- `ml_process`/`ml_process_stereo`/`ml_process2` templates (propagates to FDN reverbs, delay, resampling, generators), projections, amp/dB conversions
+  - `_core_hisstools.cpp` -- MonoConvolve, Convolver, SpectralProcessor (convolve/correlate/change_phase), KernelSmoother
+  - `_core_choc.cpp` -- FLAC read/write file I/O
+
+### Fixed
+
+- **Cross-platform build** (Linux, macOS, Windows)
+  - Linux: `CMAKE_POSITION_INDEPENDENT_CODE` for static libs linked into shared `.so`
+  - Linux: Suppressed GCC `-Wmaybe-uninitialized` false positives from HISSTools `Statistics.hpp`
+  - Linux: Dropped aarch64 wheels (HISSTools NEON code requires Apple Clang-specific implicit type conversions)
+  - macOS: Set `MACOSX_DEPLOYMENT_TARGET=10.15` for `std::filesystem::path` and nanobind aligned deallocation
+  - macOS: Architecture detection via compiler built-in defines (`__aarch64__`) instead of `CMAKE_SYSTEM_PROCESSOR` (correct under cross-compilation)
+  - macOS: `cmake/hisstools_arch_compat.h` -- bridges `__aarch64__` (GCC/Linux) to `__arm64__` (Apple/HISSTools)
+  - Windows: `NOMINMAX` and `_USE_MATH_DEFINES` for MSVC across all targets
+  - Windows: `cmake/msvc_compat.h` -- `__attribute__` no-op and `<cmath>` includes for DaisySP
+  - Python < 3.12: Guarded `AudioBuffer.__buffer__` (PEP 688) behind version check
+
+- **CI/CD** (`.github/workflows/`)
+  - `build-publish.yml` -- cibuildwheel v3.3.1 wheel builds for Linux x86_64, macOS arm64+x86_64, Windows AMD64; TestPyPI + PyPI publish via trusted publishing
+  - `ci.yml` -- QA (ruff lint/format, mypy typecheck) + native build/test matrix (ubuntu/macOS/Windows, Python 3.10+3.14)
+  - Cross-compile macOS x86_64 wheels from ARM64 runner (macos-latest); tests skipped for x86_64
+
+## [0.1.1]
+
+### Added
+
+- **CLI** (`nanodsp.__main__`, `nanodsp._cli`)
+  - `nanodsp info <file>` -- audio file metadata (path, format, duration, sample_rate, channels, frames, peak_db, loudness_lufs), `--json` output
+  - `nanodsp process <inputs...> -o OUT|-O DIR` -- chainable effect pipeline with `--fx`/`-f` (repeatable) and `--preset`/`-p` (repeatable)
+  - Batch mode: `nanodsp process *.wav -O out/` processes multiple files to an output directory
+  - Dry-run: `nanodsp process in.wav -n -f lowpass:cutoff_hz=1000` shows the chain without reading or writing files
+  - Global `-v`/`--verbose` flag for detailed step-by-step output, `-q`/`--quiet` to suppress non-essential output (mutually exclusive)
+  - `nanodsp analyze <file> <type>` -- 10 analysis subcommands (loudness, pitch, onsets, centroid, bandwidth, rolloff, flux, flatness, chromagram, info), `--json` output
+  - `nanodsp synth <out> <type>` -- 7 synth types (sine, noise, drum, oscillator, fm, note, sequence)
+  - `nanodsp convert <in> <out>` -- format conversion (WAV/FLAC), resampling (`--sample-rate`), channel conversion (`--channels`), bit depth (`-b`)
+  - `nanodsp pipe` -- read WAV from stdin, apply `-f`/`-p` effect chain, write WAV to stdout; supports Unix pipe chaining
+  - `nanodsp benchmark <function>` -- profile a DSP function with configurable iterations, warmup, buffer size; reports min/max/mean/median/std timing and realtime throughput multiplier, `--json` output
+  - `nanodsp preset list|info|apply` -- 30 presets across 8 categories (mastering, voice, spatial, dynamics, lofi, cleanup, creative)
+  - `nanodsp list [category]` -- browse all registered functions with signatures across 7 categories (filters, effects, dynamics, spectral, analysis, synthesis, ops)
+  - 13 new presets: genre mastering (`master_pop`, `master_hiphop`, `master_classical`, `master_edm`, `master_podcast`), creative effects (`radio`, `underwater`, `megaphone`, `tape_warmth`, `shimmer`, `vaporwave`, `walkie_talkie`), lofi (`8bit`)
+  - Function registry with auto-discovery from all modules, `inspect.signature`-based parameter display
+  - Preset registry with single-function and chain-based presets, parameter overrides
+  - FX token parser (`name:k=v,k=v`) with type coercion from signature defaults
+  - `[project.scripts]` entry point: `nanodsp` command
+
+- **Audio I/O** (`nanodsp.io`)
+  - `read_wav_bytes(data)` -- parse WAV from raw bytes (for stdin/pipe workflows)
+  - `write_wav_bytes(buf, bit_depth)` -- serialize AudioBuffer to WAV bytes (for stdout/pipe workflows)
+
+- **CHOC FLAC codec** -- read/write FLAC files (16/24-bit) via header-only CHOC library
+  - `nanodsp._core.choc` C++ bindings for FLAC read/write
+  - `io.read_flac()`, `io.write_flac()` Python wrappers
+  - `io.read()`, `io.write()` auto-detect WAV vs FLAC by extension
+  - Fixed CHOC upstream bug in 24-bit float-to-int scale factor
+
+- **Streaming infrastructure** (`nanodsp.stream`)
+  - `RingBuffer` -- multi-channel ring buffer with independent read/write positions
+  - `BlockProcessor` -- base class for block-based audio processors
+  - `CallbackProcessor` -- wrap a callable as a block processor
+  - `ProcessorChain` -- chain multiple processors in series
+  - `process_blocks()` -- process a buffer through a function in blocks with optional overlap-add
+
+- **DaisySP effects** (via `nanodsp.effects`)
+  - Effects: `autowah`, `chorus`, `decimator`, `flanger`, `overdrive`, `phaser`, `pitch_shift`, `sample_rate_reduce`, `tremolo`, `wavefold`, `bitcrush`, `fold`, `reverb_sc`, `dc_block`
+  - Filters: `svf_lowpass`, `svf_highpass`, `svf_bandpass`, `svf_notch`, `svf_peak`, `ladder_filter`, `moog_ladder`, `tone_lowpass`, `tone_highpass`, `modal_bandpass`, `comb_filter`
+  - Dynamics: `compress`, `limit`
+
+- **DaisySP synthesis** (via `nanodsp.synthesis`)
+  - Oscillators: `oscillator`, `fm2`, `formant_oscillator`, `bl_oscillator`
+  - Noise: `white_noise`, `clocked_noise`, `dust`
+  - Drums: `analog_bass_drum`, `analog_snare_drum`, `hihat`, `synthetic_bass_drum`, `synthetic_snare_drum`
+  - Physical modeling: `karplus_strong`, `modal_voice`, `string_voice`, `pluck`, `drip`
+
+- **STK bindings** (`nanodsp._core.stk`) -- 5 submodules, 39 classes
+  - Instruments: `Clarinet`, `Flute`, `Brass`, `Bowed`, `Plucked`, `Sitar`, `StifKarp`, `Saxofony`, `Recorder`, `BlowBotl`, `BlowHole`, `Whistle`, `Guitar`, `Twang`
+  - Generators: `SineWave`, `Noise`, `Blit`, `BlitSaw`, `BlitSquare`, `ADSR`, `Asymp`, `Envelope`, `Modulate`
+  - Filters: `BiQuad`, `OnePole`, `OneZero`, `TwoPole`, `TwoZero`, `PoleZero`, `FormSwep`
+  - Delays: `Delay`, `DelayA`, `DelayL`, `TapDelay`
+  - Effects: `FreeVerb`, `JCRev`, `NRev`, `PRCRev`, `Echo`, `Chorus`, `PitShift`, `LentPitShift`
+  - High-level wrappers: `stk_reverb`, `stk_chorus`, `stk_echo`, `synth_note`, `synth_sequence`
+
+- **Madronalib bindings** (`nanodsp._core.madronalib`) -- 7 submodules
+  - FDN reverbs: `FDN4`, `FDN8`, `FDN16` with configurable delays, cutoffs, and feedback
+  - Delays: `PitchbendableDelay`
+  - Resampling: `Downsampler`, `Upsampler`
+  - Generators: `OneShotGen`, `LinearGlide`, `SampleAccurateLinearGlide`, `TempoLock`
+  - Projections: 18 easing functions (`smoothstep`, `bell`, `ease_in`, `ease_out`, etc.)
+  - Windows: `hamming`, `blackman`, `flat_top`, `triangle`, `raised_cosine`, `rectangle`
+  - Utilities: `amp_to_db`, `db_to_amp` (scalar and array overloads)
+
+- **HISSTools bindings** (`nanodsp._core.hisstools`) -- 4 submodules
+  - Convolution: `MonoConvolve`, `Convolver` (multi-channel) with selectable latency modes
+  - Spectral processing: `SpectralProcessor` (convolve, correlate, phase change), `KernelSmoother`
+  - Analysis: 24 statistics functions (`stat_mean`, `stat_rms`, `stat_centroid`, `stat_kurtosis`, etc.), `PartialTracker`
+  - Windows: 28 window functions (Hann, Blackman-Harris variants, Nuttall variants, flat-top variants, Kaiser, Tukey, etc.)
+
+- **Spectral processing** (`nanodsp.spectral`)
+  - STFT/ISTFT with Hann window and COLA overlap-add reconstruction
+  - Spectral utilities: `magnitude`, `phase`, `from_polar`, `apply_mask`, `spectral_gate`, `spectral_emphasis`, `bin_freq`, `freq_to_bin`
+  - Spectral transforms: `time_stretch`, `phase_lock`, `spectral_freeze`, `spectral_morph`, `pitch_shift_spectral`, `spectral_denoise`
+  - `eq_match` -- match spectral envelope between two buffers
+
+- **Analysis** (`nanodsp.analysis`)
+  - Loudness: `loudness_lufs` (ITU-R BS.1770-4), `normalize_lufs`
+  - Spectral features: `spectral_centroid`, `spectral_bandwidth`, `spectral_rolloff`, `spectral_flux`, `spectral_flatness_curve`, `chromagram`
+  - Pitch detection: `pitch_detect` (YIN algorithm)
+  - Onset detection: `onset_detect` (spectral flux with peak picking)
+  - Resampling: `resample` (madronalib backend), `resample_fft` (FFT-based)
+  - Delay estimation: `gcc_phat` (GCC-PHAT)
+
+- **Composed effects** (`nanodsp.effects`)
+  - `saturate` (soft/hard/tape modes), `exciter`, `de_esser`, `parallel_compress`
+  - `noise_gate`, `stereo_delay` (with ping-pong mode), `multiband_compress`
+  - `reverb` with FDN backend and presets (room, hall, plate, chamber, cathedral)
+  - `master` -- mastering chain (dc_block, EQ, compress, limit, normalize_lufs)
+  - `vocal_chain` -- vocal processing chain (de-esser, EQ, compress, limit, normalize)
+  - `agc` -- automatic gain control with asymmetric attack/release
+
+- **Core DSP operations** (`nanodsp.ops`)
+  - Delay: `delay`, `delay_varying`
+  - Envelopes: `box_filter`, `box_stack_filter`, `peak_hold`, `peak_decay`
+  - FFT: `rfft`, `irfft`
+  - `convolve` -- FFT-based overlap-add convolution
+  - Rate conversion: `upsample_2x`, `oversample_roundtrip`
+  - Mixing: `hadamard`, `householder`, `crossfade`, `mix_buffers`
+  - `lfo` -- cubic LFO with rate/depth variation
+  - Normalization: `normalize_peak`, `trim_silence`, `fade_in`, `fade_out`
+  - Stereo: `pan`, `mid_side_encode`, `mid_side_decode`, `stereo_widen`
+  - Correlation: `xcorr` (FFT-based cross-/auto-correlation)
+  - Analytic signal: `hilbert`, `envelope`
+  - Filtering: `median_filter`, `lms_filter`
+
+- **Biquad filter wrappers** (`nanodsp.effects`)
+  - `lowpass`, `highpass`, `bandpass`, `notch`, `peak`, `peak_db`
+  - `high_shelf`, `high_shelf_db`, `low_shelf`, `low_shelf_db`, `allpass`
+  - `biquad_process` -- process through a pre-configured Biquad
+  - All accept frequency in Hz with automatic normalization
+
+- **AudioBuffer I/O methods**
+  - `AudioBuffer.from_file(path)` -- classmethod to read WAV/FLAC by extension
+  - `buf.write(path, bit_depth=16)` -- instance method to write WAV/FLAC by extension
+- `nanodsp._core.pyi` -- complete type stubs for all 12 C++ submodules
+- `Spectrogram` data class for STFT output (`[channels, frames, bins]` complex64)
+
+### Changed
+
+- **Module split** -- monolithic `dsp.py` replaced by focused modules:
+  - `_helpers.py` -- shared private utilities
+  - `ops.py` -- delay, envelopes, FFT, convolution, rates, mix, pan, normalization
+  - `effects.py` -- filters, effects, dynamics, reverb, mastering chains
+  - `spectral.py` -- STFT, spectral utilities, spectral transforms, eq_match
+  - `synthesis.py` -- oscillators, noise, drums, physical modeling, STK synth
+  - `analysis.py` -- loudness, spectral features, pitch/onset detection, resampling
+- `__init__.py` stripped to `__version__` only -- no re-exports; use explicit imports
+- `io.py` now supports both WAV and FLAC formats
+- Test suite reorganized into per-module test files (1114 tests)
+- Removed `disable_error_code = ["import-untyped"]` from mypy config (stubs fix this)
+
+## [0.1.0]
+
+### Added
+
+- Initial project structure with scikit-build-core + CMake + uv
+- Core C++ bindings via nanobind (`nanodsp._core`):
+  - `filters` -- `Biquad` with 16 filter designs, `BiquadDesign` enum
+  - `fft` -- `FFT` (complex-to-complex), `RealFFT` (real-to-complex)
+  - `delay` -- `Delay` (linear interpolation), `DelayCubic` (cubic interpolation)
+  - `envelopes` -- `CubicLfo`, `BoxFilter`, `BoxStackFilter`, `PeakHold`, `PeakDecayLinear`
+  - `spectral` -- `STFT` (multi-channel analysis/synthesis)
+  - `rates` -- `Oversampler2x`
+  - `mix` -- `Hadamard`, `Householder`, `cheap_energy_crossfade`
+- `AudioBuffer` class (pure Python, 2D `[channels, frames]` float32 with metadata)
+  - Factory methods: `zeros`, `ones`, `impulse`, `sine`, `noise`, `from_numpy`
+  - Channel operations: `to_mono`, `to_channels`, `split`, `concat_channels`
+  - Arithmetic operators: `+`, `-`, `*`, `/`, `gain_db`
+  - Pipeline: `pipe()` for chaining DSP functions
+- `io.read_wav()`, `io.write_wav()` -- WAV file I/O (8/16/24/32-bit PCM, stdlib `wave`)
+- Test suite with pytest (203 tests)