Files
moloapp/src/google/gemini_voice.h
T
retoorandClaude Sonnet 5.5 837de58136 feat(clients): google cloud voice: generic client, TTS, STT, Gemini audio, live voice page
src/google: GoogleClient over molo::clients (key in x-goog-api-key, never in
URLs), Service base, TextToSpeech, SpeechToText, Generative (replies, audio
in, audio out), Recognizer/Synthesizer seams with cloud, gemini and vertex
backends, VoicePipeline with bounded history. Example: /voice live
conversation page, /voice/{say,hear,turn,reset}, /google/status. Keys load
from env, .env.json.local, .env.json. Paid routes are excluded from locust.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-24 04:05:41 +02:00

38 lines
1.0 KiB
C++

#pragma once
// Recognizer and Synthesizer over Gemini: audio understanding for speech to
// text, Gemini TTS models for text to speech. Works with any credential the
// Gemini endpoint accepts (AI Studio key, Vertex express key, OAuth).
#include <memory>
#include <string>
#include "google/generative.h"
#include "google/voice_io.h"
namespace molo::google {
class GeminiRecognizer : public Recognizer {
public:
explicit GeminiRecognizer(std::shared_ptr<Generative> generative,
std::string model = "");
Transcript recognize(const RecognitionRequest& request) const override;
private:
std::shared_ptr<Generative> generative_;
std::string model_;
};
class GeminiSynthesizer : public Synthesizer {
public:
explicit GeminiSynthesizer(std::shared_ptr<Generative> generative,
std::string model = "");
Audio synthesize(const SynthesisRequest& request) const override;
private:
std::shared_ptr<Generative> generative_;
std::string model_;
};
} // namespace molo::google