pi-voicekit 0.2.2 → 0.2.3

package/README.md

@@ -401,7 +401,8 @@ these entries — they are not part of the conversation context — so they are
 analysis, and they do keep the raw text on disk for as long as the session file exists.
 Each entry also records how the pass was configured: the transcript length on its own
 (separate from any text already in the editor), whether thinking was turned off for it,
-and the output-token cap it carried.
+and the output-token cap it carried, plus the audio seconds it covered and which recogniser
+produced it — which is what makes polish time readable as a speedup.
 
 | Setting                   | Scope              | Default     | Notes                                                   |
 | ------------------------- | ------------------ | ----------- | ------------------------------------------------------- |

package/extensions/voice/post-process.ts

@@ -244,6 +244,10 @@ export interface PolishAudit {
 	transcriptChars?: number;
 	/** Characters of the editor prefix, which is not part of the transcript. */
 	editorPrefixChars?: number;
+	/** Seconds of audio the dictation covered, so polish time can be read as a speedup. */
+	durationSec?: number;
+	/** Which recogniser produced it, so results are never pooled across backends. */
+	backend?: string;
 	/** True when the pass asked the model not to think. */
 	thinkingOff?: boolean;
 	/** The output-token cap the request carried, a cap and not a spend. */
@@ -259,6 +263,8 @@ export function buildPolishAudit(input: {
 	transcriptChars?: number;
 	editorPrefixChars?: number;
 	thinkingOff?: boolean;
+	durationSec?: number;
+	backend?: string;
 	maxTokens?: number;
 	telemetry?: {
 		model?: string;
@@ -282,6 +288,8 @@ export function buildPolishAudit(input: {
 	if (input.editorPrefixChars !== undefined) audit.editorPrefixChars = input.editorPrefixChars;
 	if (input.thinkingOff !== undefined) audit.thinkingOff = input.thinkingOff;
 	if (input.maxTokens !== undefined) audit.maxTokens = input.maxTokens;
+	if (input.durationSec !== undefined) audit.durationSec = input.durationSec;
+	if (input.backend !== undefined) audit.backend = input.backend;
 	const telemetry = input.telemetry;
 	if (telemetry) {
 		if (telemetry.model !== undefined) audit.model = telemetry.model;

package/extensions/voice.ts

@@ -1830,6 +1830,8 @@ export default function (pi: ExtensionAPI) {
 									editorPrefixChars: prefix.length,
 									thinkingOff: polishTelemetry?.thinkingOff,
 									maxTokens: polishTelemetry?.maxTokens,
+									durationSec: Number(elapsed),
+									backend: config.backend,
 									written: wroteEditor ? finalText : undefined,
 									status: final.status,
 									disposition: final.disposition,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-voicekit",
-  "version": "0.2.2",
+  "version": "0.2.3",
   "description": "Voice in + voice out for Pi CLI — hold-to-talk STT (Deepgram streaming or 21 offline models) plus TTS (Kitten Nano, Piper, Kokoro, or Deepgram Aura)",
   "type": "module",
   "keywords": [