#include "vendors/audio/deepgram/speak.h" namespace molo::audio::deepgram { using C = DeepgramCalls; Json Speak::build(const SpeakOptions& o) const { require_text("text", o.text, DeepgramCaps::kMaxSpeakChars); Json body = Json::object(); body.set("text", Json(o.text)); return body; } Query Speak::query(const SpeakOptions& o, bool v2) const { Query q; if (!o.model.empty()) { if (!valid_token(o.model)) bad("model has characters Deepgram does not accept"); q.add("model", o.model); } if (!o.encoding.empty()) { if (!valid_token(o.encoding)) bad("bad encoding"); q.add("encoding", o.encoding); } if (!o.container.empty()) { if (!valid_token(o.container)) bad("bad container"); q.add("container", o.container); } if (o.sample_rate < 0 || o.sample_rate > 96000) bad("sample_rate must be 0 or up to 96000"); if (o.sample_rate > 0) q.add_int("sample_rate", o.sample_rate); if (o.bit_rate < 0 || o.bit_rate > 1000000) bad("bit_rate is out of range"); if (o.bit_rate > 0) q.add_int("bit_rate", o.bit_rate); if (o.speed < 0) bad("speed must not be negative"); if (o.speed > 0) q.add("speed", std::to_string(o.speed)); if (v2) { q.add_if("expressivity", o.expressivity); q.add_if("priority", o.priority); } else if (!o.expressivity.empty() || !o.priority.empty()) { bad("expressivity and priority belong to /v2/speak"); } q.add_if("tag", o.tag); q.add_if("callback", o.callback); q.add_if("callback_method", o.callback_method); if (o.mip_opt_out >= 0) q.add_bool("mip_opt_out", o.mip_opt_out == 1); check_query(table_.get(v2 ? C::kSpeakV2 : C::kSpeak), q); return q; } SpeakResult Speak::synthesize(const SpeakOptions& o, bool v2) { Json body = build(o); llm::RawResponse raw = call_raw(v2 ? C::kSpeakV2 : C::kSpeak, {}, &body, query(o, v2)); return SpeakResult{raw.body, raw.content_type}; } ByteStreamer::Result Speak::stream(const SpeakOptions& o, const ChunkSink& sink, bool v2) { Json body = build(o); return call_stream(v2 ? C::kSpeakV2 : C::kSpeak, {}, &body, query(o, v2), sink); } } // namespace molo::audio::deepgram