Files
moloapp/src/vendors/audio/elevenlabs/stt.cpp
T
retoorandClaude Sonnet 5.5 6d80dd0223 feat(clients): audio: ElevenLabs, AssemblyAI, Deepgram, Cartesia, Speechmatics, Hume
Table driven REST (509 rows verified against vendor specs by tools/check_spec.py),
WebSockets over the src/google/live client, neutral SpeechSynthesizer /
SpeechRecognizer seam, /ai/audio demo with dry runs, 110 package tests and 5
route tests. PlayHT has no public API; LeMUR is gone from AssemblyAI's spec.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
2026-10-25 15:36:57 +01:00

67 lines
3.0 KiB
C++

#include "vendors/audio/elevenlabs/stt.h"
#include "vendors/audio/caps.h"
namespace molo::audio::elevenlabs {
using C = ElevenCalls;
clients::body::MultipartBody SpeechToText::build(const SttRequest& r) const {
require_text("model_id", r.model_id, 128);
int sources = (!r.audio.empty()) + (!r.source_url.empty()) + (!r.cloud_storage_url.empty());
if (sources != 1) bad("give exactly one of audio, source_url and cloud_storage_url");
if (r.audio.size() > ElevenCaps::kMaxFileBytes) bad("audio is over the byte cap");
require_text("source_url", r.source_url, AudioCaps::kMaxFieldChars, false);
require_text("cloud_storage_url", r.cloud_storage_url, AudioCaps::kMaxFieldChars, false);
require_text("language_code", r.language_code, 16, false);
if (r.num_speakers < 0 || r.num_speakers > 32) bad("num_speakers must be 0 or 1..32");
const std::string& g = r.timestamps_granularity;
if (!g.empty() && g != "none" && g != "word" && g != "character")
bad("timestamps_granularity must be none, word or character");
if (!r.file_format.empty() && r.file_format != "pcm_s16le_16" && r.file_format != "other")
bad("file_format must be pcm_s16le_16 or other");
if (r.keyterms.size() > AudioCaps::kMaxKeywords) bad("too many keyterms");
if (r.temperature > 2.0) bad("temperature must be 0..2");
clients::body::MultipartBody form;
form.add_text("model_id", r.model_id);
if (!r.audio.empty()) form.add_file("file", r.filename, r.content_type, r.audio);
if (!r.source_url.empty()) form.add_text("source_url", r.source_url);
if (!r.cloud_storage_url.empty()) form.add_text("cloud_storage_url", r.cloud_storage_url);
if (!r.language_code.empty()) form.add_text("language_code", r.language_code);
if (r.diarize) form.add_text("diarize", "true");
if (r.num_speakers > 0) form.add_text("num_speakers", std::to_string(r.num_speakers));
if (!g.empty()) form.add_text("timestamps_granularity", g);
if (r.tag_audio_events >= 0) form.add_text("tag_audio_events", r.tag_audio_events ? "true" : "false");
for (const auto& k : r.keyterms) {
require_text("keyterm", k, 100);
form.add_text("keyterms", k);
}
if (!r.file_format.empty()) form.add_text("file_format", r.file_format);
if (r.use_multi_channel >= 0) form.add_text("use_multi_channel", r.use_multi_channel ? "true" : "false");
if (r.seed >= 0) form.add_text("seed", std::to_string(r.seed));
if (r.temperature >= 0) form.add_text("temperature", std::to_string(r.temperature));
return form;
}
Query SpeechToText::query(const SttRequest& r) const {
Query q;
if (r.enable_logging >= 0) q.add_bool("enable_logging", r.enable_logging == 1);
return q;
}
Json SpeechToText::transcribe(const SttRequest& r) {
return call_form(C::kStt, {}, build(r), query(r));
}
Json SpeechToText::get(const std::string& id) {
check_id("transcription", id);
return call(C::kSttGet, {id});
}
Json SpeechToText::remove(const std::string& id, bool confirm) {
check_id("transcription", id);
return call_destructive(C::kSttDelete, {id}, nullptr, Query(), confirm);
}
} // namespace molo::audio::elevenlabs