audiobits 0.1.1 → 0.1.3

package/CHANGELOG.md

@@ -1,5 +1,18 @@
 # audiobits
 
+## 0.1.3
+
+### Patch Changes
+
+- 7aa39bd: Align the package README with published capabilities and add the public website
+  and relevant npm keywords to package metadata. Runtime behavior is unchanged.
+
+## 0.1.2
+
+### Patch Changes
+
+- 7f792f9: Clarify npm installation and the minimal confirmation-sound quick start in the shipped README and Skill, with nearby browser gesture and lifecycle guidance.
+
 ## 0.1.1
 
 ### Patch Changes

package/README.md

@@ -1,20 +1,23 @@
 # AudioBits
 
+[Website and documentation](https://audiobits.joacod.com/)
+
 AudioBits is available on npm as `audiobits`. It includes confirmation, impact, thruster, tactile click, gentle rejection,
 glass notification, whoosh and power-up; schema-1 recipes; play/live controls; seeded variation; bounded
 voices; bus gain, mute and routing; native output taps; and explicit lifecycle APIs.
-Prerelease verification target: current Chromium. Other browsers and operating
-systems are unverified and deliberately deferred. Maintainer listening acceptance for the
+Current Chromium is the only automated browser verification target. Other
+browsers and operating systems are unverified and deliberately deferred. Maintainer listening acceptance for the
 unchanged sound definitions is recorded separately.
 
-## Install
+## Quick start
 
 ```sh
 npm install audiobits
 ```
 
-Use a browser bundler and call `play()` directly from a gesture handler. Surface rejection
-with `void play().catch(showError)` and retry from a fresh gesture.
+Use a browser bundler. Create the engine and sound once; run the final two lines
+inside a user gesture handler, such as a button click. Call `audio.start()`
+directly in that handler and await it before playback.
 
 ```ts
 import { createAudio } from "audiobits";
@@ -22,19 +25,15 @@ import { confirmation } from "audiobits/recipes";
 
 const audio = createAudio();
 const sound = audio.sound(confirmation);
-export async function play() {
-  await audio.start();
-  sound.play();
-}
-export function stop() {
-  audio.stopAll({ tails: "cut" });
-}
-export async function dispose() {
-  await audio.dispose();
-}
+
+await audio.start();
+sound.play();
 ```
 
-Bind `play()` to a button click (`void play().catch(showError)`).
+Surface activation failures and retry from a fresh gesture. Use
+`audio.stopAll({ tails: "cut" })` to stop playback and `await audio.dispose()`
+when the host is removed. See the production lifecycle example below for
+hiding and navigation.
 
 ## Production lifecycle
 
@@ -116,7 +115,7 @@ Schema version 1 supports:
 Depth is limited to 16, visited values to 10000, and diagnostics to 100.
 Source/filter frequencies must be below the owning context's Nyquist frequency
 at playback, including all mapping/variation extrema. Other effects,
-expression strings, sequencing, and continuous random modulation are rejected. These are the capabilities shipped in 0.1.0.
+expression strings, sequencing, and continuous random modulation are rejected.
 The generated schema is available as `recipeSchema` or `audiobits/schema.json`.
 
 ## Runtime ownership
@@ -213,21 +212,14 @@ promise identical oscillator/filter samples across browsers or sample rates.
 
 ## Noise ownership and bounds
 
-Each noise layer owns one mono, one-second buffer at context sample rate. The
-last 20 ms crossfades into the first 20 ms; looping resumes after that prefix,
-so the wrap follows an ordinary adjacent sample pair. The loop period is about
-0.98 seconds. Both finite and sustained noise loop this bounded resource.
-This treatment has signal evidence and maintainer listening acceptance;
-individual listening observations were not recorded. There is no shared noise cache or rendered-output cache.
+Each noise layer uses a bounded generated mono buffer at the context sample rate,
+looped with a crossfade for finite and sustained playback. Noise generation is
+reproducible for the same recipe, controls, seed and sample rate. There is no
+shared noise cache or rendered-output cache.
 
 Noise supports integer sample rates from 8000 through 192000 Hz. Unsupported
-rates fail with `noise-rate` before voice allocation. A layer retains at most
-768000 sample bytes (192000 at 48 kHz); generation temporarily uses one extra
-buffer of that size. A recipe has at most 16 layers, and existing voice limits
-bound concurrent buffers. The thruster owns ten nodes and one buffer per voice;
-with its default eight-voice limit plus one retiree, sample storage is at most
-1728000 bytes at 48 kHz. The engine master owns two additional nodes for independent gain and mute. These are
-owned sample-storage bounds, not measurements of all browser memory.
+rates fail with `noise-rate` before voice allocation. Recipe layer limits and
+engine/sound voice limits bound concurrent noise resources.
 
 Control updates allocate no new audio nodes or buffers. Finishing a voice clears
 its buffer-source references and disconnects owned nodes, including cancellation
@@ -312,9 +304,8 @@ and suspend on hide. Returning does not resume; a fresh gesture starts audio.
 Late activation after suspension is cancelled and cannot replay an old request.
 The engine also clears owned voices/effects on native interruption; automatic
 native recovery is suspended until a fresh `start()` request. Physical OS
-interruption still requires manual evidence. Step 04 passed maintainer manual
-verification on 2026-10-06. Browser/device details and individual observations
-were not supplied; automated checks remain separate from that acceptance.
+interruption still requires manual evidence. Automated checks do not establish
+physical-device compatibility.
 
 Buses provide gain, mute, and routing. Shared effects are deferred until
 multiple real sound requirements justify an API.

package/dist/capabilities.json

@@ -1,5 +1,5 @@
 {
-  "packageVersion": "0.1.1",
+  "packageVersion": "0.1.3",
   "schemaVersion": 1,
   "status": "stable",
   "recipe": {

package/package.json

@@ -1,8 +1,17 @@
 {
   "name": "audiobits",
-  "version": "0.1.1",
+  "version": "0.1.3",
   "description": "Procedural browser audio: versioned recipes, live controls, managed voices and buses.",
   "license": "MIT",
+  "homepage": "https://audiobits.joacod.com/",
+  "keywords": [
+    "web-audio",
+    "procedural-audio",
+    "sound-effects",
+    "sfx",
+    "game-audio",
+    "typescript"
+  ],
   "repository": {
     "type": "git",
     "url": "https://github.com/joacod/audiobits.git",

package/skill/SKILL.md

@@ -8,10 +8,13 @@ description: Integrate AudioBits procedural sounds into a browser application, i
 This guide ships with the `audiobits` npm package. Confirm the installed
 package version and read `audiobits/capabilities.json` and
 `audiobits/schema.json` before choosing primitives. Package metadata describes
-shipped behavior, not registry availability. Prerelease verification target: current Chromium. Other browsers and operating
-systems are unverified and deliberately deferred.
+shipped behavior, not registry availability. Current Chromium is the only
+automated browser verification target. Other browsers and operating systems are
+unverified and deliberately deferred.
 
-Use public imports from `audiobits` and `audiobits/recipes`. Create engines and
+Install with `npm install audiobits` and use a browser bundler. Follow the
+[package quick start](../README.md#quick-start) to play the bundled `confirmation`
+recipe with public imports from `audiobits` and `audiobits/recipes`. Create engines and
 sounds without browser activation, then invoke `start()` synchronously inside
 a user gesture and await it before `play()`. Surface failures and allow a fresh
 gesture retry. Never queue playback across Stop, hiding, or disposal.