// Route: POST /voice/turn (Gemini audio backend: Vertex express key) #include "support/GoogleRig.hpp" using namespace molo::example::test; int main() { setenv("GOOGLE_GENAI_MODEL", "chat-m", 1); setenv("GOOGLE_STT_MODEL", "stt-m", 1); setenv("GOOGLE_TTS_MODEL", "tts-m", 1); GoogleRig rig; const std::string p = "/v1/publishers/google/models/"; auto text = [&](const std::string& model, const std::string& t) { rig.mock->expect_response( molo::clients::Method::kPost, p + model + ":generateContent", GoogleRig::json("{\"candidates\":[{\"content\":{\"parts\":[{\"text\":\"" + t + "\"}]}}]}")); }; text("stt-m", "hallo daar"); text("chat-m", "Hoi!"); rig.mock->expect_response( molo::clients::Method::kPost, p + "tts-m:generateContent", GoogleRig::json( "{\"candidates\":[{\"content\":{\"parts\":[{\"inlineData\":{\"mimeType\":" "\"audio/L16;codec=pcm;rate=24000\",\"data\":\"" + molo::security::crypto::Base64::encode("PCM") + "\"}}]}}]}")); Fixture f(rig.client, molo::google::VoiceBackend::kVertex); check(contains(f.client->get("/google/status").body, "\"backend\":\"vertex\""), "status names the backend"); auto res = f.client->post("/voice/turn?cid=v1", "WEBM", "audio/webm;codecs=opus"); check(res.status == 200, "status 200"); check(contains(res.body, "\"transcript\":\"hallo daar\""), "transcript"); check(contains(res.body, "\"reply\":\"Hoi!\""), "reply"); check(contains(res.body, "\"mime\":\"audio/wav\""), "wav for the browser"); check(rig.mock->send_count() == 3, "three Gemini calls"); check(rig.mock->seen()[0].target.rfind(p, 0) == 0, "vertex endpoint used"); check(contains(rig.mock->seen()[0].body, "audio/webm"), "audio mime sent to Gemini"); check(rig.mock->seen()[0].headers.get_or("x-goog-api-key", "") == kTestKey, "key in header"); unsetenv("GOOGLE_GENAI_MODEL"); unsetenv("GOOGLE_STT_MODEL"); unsetenv("GOOGLE_TTS_MODEL"); return finish("voice/turn vertex"); }