Add in-memory voice-note codec helpers

.gitignore

@@ -26,3 +26,4 @@ worktrees/*
 .idea/
 .vscode/
 *.swp
+/live-module-coverage.json

MODULE.md

@@ -4,12 +4,15 @@ This package contains the native `tinyvoice` module for TinyBus module ABI v1.
 Install only the archive matching the host operating system and architecture.
 
 The module claims `ai.tinyhumans.tinyvoice.Voice`, serves the object at
-`/ai/tinyhumans/tinyvoice/Voice`, and provides 53 methods under contract 1.10.
+`/ai/tinyhumans/tinyvoice/Voice`, and provides 56 methods under contract 1.11.
 
 ## Methods
 
 | Method | Arguments | Returns |
 | --- | --- | --- |
+| `CodecProbe` | `CodecInput { audio_base64 }` | `CodecInfo`: detected codec, duration, sample rate, channels |
+| `CodecTranscode` | `CodecTranscodeRequest { audio_base64, from, target }` | `CodecOutput { audio_base64 }` |
+| `CodecWaveform` | `WaveformRequest { audio_base64, buckets }` | `WaveformResult { peaks }` |
 | `Route` | `transcript` | JSON `VoiceIntent`, tagged by an `intent` field |
 | `ExtractCommand` | `transcript`, `wake_word` | The command after the wake word, or `""` |
 | `WakeWordPresent` | `transcript`, `wake_word` | `bool` |

README.md

@@ -1,6 +1,6 @@
 # TinyVoice
 
-Host-agnostic voice primitives, extracted from [OpenHuman](https://github.com/tinyhumansai/openhuman): WAV framing, silence gating, voice-activity segmentation, wake-word gating, fast-path command routing, STT hallucination detection, and relay-turn preparation.
+Host-agnostic voice primitives, extracted from [OpenHuman](https://github.com/tinyhumansai/openhuman): WAV framing and voice-note codecs, silence gating, voice-activity segmentation, wake-word gating, fast-path command routing, STT hallucination detection, and relay-turn preparation.
 
 Ships two things from one repository:
 
@@ -25,6 +25,7 @@ permission check.
 | Here | With the host |
 | --- | --- |
 | WAV framing, RMS, resampling, downmix, silence gate; module-owned capture threads | Microphone permission decisions through the computer module |
+| In-memory WAV/PCM16, Ogg/Opus, and M4A/AAC-LC probe, transcode, and waveform APIs | Codec calls through the `tinyvoice-bus` module contract |
 | VAD segmentation | The capture loop that drives it |
 | Wake-word gate, intent routing | What to *do* with an intent |
 | Hallucination detection | The STT transport, credentials, and retry policy |
@@ -66,6 +67,11 @@ turn. It performs no inference or tool execution. A module host uses the
 OpenHuman supplies its local agent, acknowledgment policy and delivery.
 See the [relay contract](docs/specs/relay-turn-preparation.md).
 
+The [`tinyvoice::codec` API](docs/specs/voice-codecs.md) detects, probes,
+transcodes, and extracts waveform peaks from supported voice-note formats. It
+uses pure-Rust codecs and in-memory containers; it does not require FFmpeg or
+system codec libraries.
+
 ### Optional features
 
 - `hotkey` (off by default): `tinyvoice::hotkey`, a system-wide push-to-talk /