[P1] Export WAV, samples, coefficients, and executable signal functions #13

Parent plan: #1
Priority / gate: P1 / M2
Requirement: F09 in the PRD
Design: Technical design
Depends on: #5, #6, #7, #8, #11, #9

Outcome

Let users take the audible result and its mathematical definition into other tools with enough metadata to reproduce them.

Scope

• Expose a shared finite pre-monitor render for soundboard caching. Per-sound WAV/project downloads use the saved snapshot and existing adapters; apply the chosen master monitor gain once at WAV output, without baking it into the cached clip.
• Encode mono 16-bit PCM WAV at 44.1/48 kHz; periodic export uses a finite duration, default 2 seconds and maximum 10 seconds.
• Add Source/Rendered CSV selection, complete coefficient JSON, and JavaScript/Python evaluators with data and usage examples.
• Share the rendering contract with preview; rendered WAV includes selected source, conditioning, monitor gain, and finite start/end envelopes.
• Identify source, sine fit, Fourier approximation, and rendered-output function exports explicitly.
• Preserve computational precision and attach schema/algorithm versions, units, rates, durations, and processing settings.
• Export mixed or selected-oscillator audio; include all voices' identity, carrier/model, phase, pitch/gain curves, interpolation, inclusion/mute/solo state, and global attenuation metadata. Generated functions include per-voice phase integrals and the complete sum.
• Expose export adapters for manifest/companion sample assets and resettable stateful evaluators. [P1] Import audio and preserve reusable clip assets in portable projects #19 owns complete asset bundle integration; [P1] Modulate sounds with imported audio or saved clips #20 owns modulation formula/function/audio export integration and its numerical tests.

Acceptance criteria

• A saved-sound pre-monitor buffer and its monitored WAV reproduce the reference render at matched settings; a non-unity monitor-gain fixture detects accidental double attenuation.
• Decoded WAV headers, sample count, rate, channels, and samples agree with the expected rendered buffer within one PCM quantization step.
• Source JS/Python evaluators match fixtures within 10^-6, including boundaries and periodic wrap; rendered evaluators reproduce their exported numerical waveform.
• Changing export rate recomputes bandwidth/rendering at that rate and is visible before download.
• Complex signals include the full required table/coefficients; the UI never exports a truncated pretty-printed formula as a complete definition.
• CSV labels seconds and amplitude and identifies Source or Rendered; JSON exports contain finite values and sufficient reproduction settings.
• Exporting does not mutate the editor, and failures/cancellation preserve the active project.
• Constant envelopes reproduce the equivalent oscillator output; JS/Python phase-integrated sweeps and mixed evaluators match reference fixtures, and intentional gain dynamics are retained.

Validation

Independent WAV decode checks, positive/negative PCM extrema, CSV/JSON schema checks, execution of generated JS/Python fixtures, and preview/offline DSP comparisons at matched settings.

Boundaries

Compressed audio encoding and bit-identical device output are deferred. Local WAV/MP3 imports are included in #19; no server upload is required.

Implementation evidence

WAV, CSV representations, complete coefficients and executable source/modulated/rendered JS/Python functions are implemented with rate/scope selection and snapshot/cancel behavior. Independent language execution and browser downloads pass.

Merged work: #25, #28, #29.

Validated CI run · Validation report. The report arrives with the final documentation PR; numerical/browser logs already exist. Release-wide human acceptance and the recorded performance/memory gaps remain explicit in #15 and #16.