Shifty/Mastermind
Shifty/Mastermind
1
1
Fork 0
Code Issues 1 Pull Requests Actions Packages Projects 1 Releases Wiki Activity

Add logging for transcription handling and disable proactive audio

Browse Source
This commit is contained in:
Илья Глазунов 2026-02-14 04:28:29 +03:00
parent 310b6b3fbd
commit bfd76dc0c1
1 changed files with 11 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
src/utils/gemini.js
View File

@ -473,10 +473,13 @@ async function initializeGeminiSession(apiKey, customPrompt = '', profile = 'int
// Handle input transcription (what was spoken) // Handle input transcription (what was spoken)
if (message.serverContent?.inputTranscription?.results) { if (message.serverContent?.inputTranscription?.results) {
currentTranscription += formatSpeakerResults(message.serverContent.inputTranscription.results); const transcribed = formatSpeakerResults(message.serverContent.inputTranscription.results);
console.log('Got transcription (results):', transcribed);
currentTranscription += transcribed;
} else if (message.serverContent?.inputTranscription?.text) { } else if (message.serverContent?.inputTranscription?.text) {
const text = message.serverContent.inputTranscription.text; const text = message.serverContent.inputTranscription.text;
if (text.trim() !== '') { if (text.trim() !== '') {
console.log('Got transcription (text):', text);
currentTranscription += text; currentTranscription += text;
} }
} }
@ -485,18 +488,23 @@ async function initializeGeminiSession(apiKey, customPrompt = '', profile = 'int
// if (message.serverContent?.outputTranscription?.text) { ... } // if (message.serverContent?.outputTranscription?.text) { ... }
if (message.serverContent?.generationComplete) { if (message.serverContent?.generationComplete) {
console.log('Generation complete. Current transcription:', `"${currentTranscription}"`);
if (currentTranscription.trim() !== '') { if (currentTranscription.trim() !== '') {
console.log('Sending to', hasGroqKey() ? 'Groq' : 'Gemma');
if (hasGroqKey()) { if (hasGroqKey()) {
sendToGroq(currentTranscription); sendToGroq(currentTranscription);
} else { } else {
sendToGemma(currentTranscription); sendToGemma(currentTranscription);
} }
currentTranscription = ''; currentTranscription = '';
} else {
console.log('Transcription is empty, not sending to LLM');
} }
messageBuffer = ''; messageBuffer = '';
} }
if (message.serverContent?.turnComplete) { if (message.serverContent?.turnComplete) {
console.log('Turn complete');
sendToRenderer('update-status', 'Listening...'); sendToRenderer('update-status', 'Listening...');
} }
}, },
@ -524,15 +532,10 @@ async function initializeGeminiSession(apiKey, customPrompt = '', profile = 'int
}, },
config: { config: {
responseModalities: [Modality.AUDIO], responseModalities: [Modality.AUDIO],
proactivity: { proactiveAudio: true }, proactivity: { proactiveAudio: false },
outputAudioTranscription: {}, outputAudioTranscription: {},
inputAudioTranscription: {},
tools: enabledTools, tools: enabledTools,
// Enable speaker diarization
// inputAudioTranscription: {
// enableSpeakerDiarization: true,
// minSpeakerCount: 2,
// maxSpeakerCount: 2,
// },
contextWindowCompression: { slidingWindow: {} }, contextWindowCompression: { slidingWindow: {} },
speechConfig: { languageCode: language }, speechConfig: { languageCode: language },
systemInstruction: { systemInstruction: {