speech-core/src/speech_core_c.cpp at main · soniqo/speech-core · GitHub

History

433 lines (367 loc) · 14.7 KB

Raw

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

381

382

383

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

414

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

430

431

432

433

#include "speech_core/speech_core_c.h"

#include "speech_core/pipeline/voice_pipeline.h"

#include <memory>

#include <string>

#include <vector>

using namespace speech_core;

// ---------------------------------------------------------------------------

// Adapter classes — bridge C vtable function pointers to C++ interfaces

// ---------------------------------------------------------------------------

class CSTTAdapter : public STTInterface {

sc_stt_vtable_t vt_;

std::string last_partial_; // keeps C string alive between push_chunk calls

public:

explicit CSTTAdapter(sc_stt_vtable_t vt) : vt_(vt) {}

TranscriptionResult transcribe(

const float* audio, size_t length, int sample_rate) override

{

auto r = vt_.transcribe(vt_.context, audio, length, sample_rate);

return {

r.text ? std::string(r.text) : "",

r.language ? std::string(r.language) : "",

r.confidence,

r.start_time,

r.end_time

};

}

int input_sample_rate() const override {

return vt_.input_sample_rate(vt_.context);

}

bool supports_streaming() const override {

return vt_.begin_stream != nullptr;

}

void begin_stream(int sample_rate) override {

if (vt_.begin_stream) vt_.begin_stream(vt_.context, sample_rate);

}

PartialResult push_chunk(const float* audio, size_t length) override {

if (!vt_.push_chunk) return {};

auto r = vt_.push_chunk(vt_.context, audio, length);

return {

r.text ? std::string(r.text) : "",

r.language ? std::string(r.language) : "",

r.confidence

};

}

void flush_stream() override {

if (vt_.flush_stream) vt_.flush_stream(vt_.context);

}

TranscriptionResult end_stream() override {

if (!vt_.end_stream) return {};

auto r = vt_.end_stream(vt_.context);

return {

r.text ? std::string(r.text) : "",

r.language ? std::string(r.language) : "",

r.confidence, r.start_time, r.end_time

};

}

void cancel_stream() override {

if (vt_.cancel_stream) vt_.cancel_stream(vt_.context);

}

};

class CTTSAdapter : public TTSInterface {

sc_tts_vtable_t vt_;

public:

explicit CTTSAdapter(sc_tts_vtable_t vt) : vt_(vt) {}

void synthesize(const std::string& text, const std::string& language,

TTSChunkCallback on_chunk) override

{

// Bridge C++ std::function to C function pointer + context

vt_.synthesize(vt_.context, text.c_str(), language.c_str(),

[](const float* samples, size_t length, bool is_final, void* ctx) {

auto* fn = static_cast<TTSChunkCallback*>(ctx);

(*fn)(samples, length, is_final);

&on_chunk);

}

int output_sample_rate() const override {

return vt_.output_sample_rate(vt_.context);

}

void cancel() override {

if (vt_.cancel) vt_.cancel(vt_.context);

}

};

class CVADAdapter : public VADInterface {

sc_vad_vtable_t vt_;

public:

explicit CVADAdapter(sc_vad_vtable_t vt) : vt_(vt) {}

float process_chunk(const float* samples, size_t length) override {

return vt_.process_chunk(vt_.context, samples, length);

}

void reset() override {

vt_.reset(vt_.context);

}

int input_sample_rate() const override {

return vt_.input_sample_rate(vt_.context);

}

size_t chunk_size() const override {

return vt_.chunk_size(vt_.context);

}

};

class CLLMAdapter : public LLMInterface {

sc_llm_vtable_t vt_;

public:

explicit CLLMAdapter(sc_llm_vtable_t vt) : vt_(vt) {}

LLMResponse chat(const std::vector<Message>& messages,

LLMTokenCallback on_token) override

{

// Convert C++ messages to C array

std::vector<sc_message_t> c_msgs(messages.size());

for (size_t i = 0; i < messages.size(); i++) {

c_msgs[i].role = static_cast<sc_role_t>(messages[i].role);

c_msgs[i].content = messages[i].content.c_str();

}

// Bridge C++ std::function to C function pointer + context

vt_.chat(vt_.context, c_msgs.data(), c_msgs.size(),

[](const char* token, bool is_final, void* ctx) {

auto* fn = static_cast<LLMTokenCallback*>(ctx);

(*fn)(std::string(token), is_final);

&on_token);

return {}; // C API doesn't support tool calls yet

}

void cancel() override {

if (vt_.cancel) vt_.cancel(vt_.context);

}

};

class CEnhancerAdapter : public EnhancerInterface {

sc_enhancer_vtable_t vt_;

public:

explicit CEnhancerAdapter(sc_enhancer_vtable_t vt) : vt_(vt) {}

void enhance(const float* audio, size_t length, int sample_rate,

float* output) override {

vt_.enhance(vt_.context, audio, length, sample_rate, output);

}

int input_sample_rate() const override {

return vt_.input_sample_rate(vt_.context);

}

};

class CEchoCancellerAdapter : public EchoCancellerInterface {

sc_echo_canceller_vtable_t vt_;

public:

explicit CEchoCancellerAdapter(sc_echo_canceller_vtable_t vt) : vt_(vt) {}

void feed_reference(const float* samples, size_t length) override {

vt_.feed_reference(vt_.context, samples, length);

}

void cancel_echo(const float* input, size_t length, float* output) override {

vt_.cancel_echo(vt_.context, input, length, output);

}

int input_sample_rate() const override {

return vt_.input_sample_rate(vt_.context);

}

void reset() override {

if (vt_.reset) vt_.reset(vt_.context);

}

};

// ---------------------------------------------------------------------------

// Pipeline handle

// ---------------------------------------------------------------------------

struct sc_pipeline_s {

std::unique_ptr<CSTTAdapter> stt;

std::unique_ptr<CTTSAdapter> tts;

std::unique_ptr<CLLMAdapter> llm;

std::unique_ptr<CVADAdapter> vad;

std::unique_ptr<CEnhancerAdapter> enhancer;

std::unique_ptr<CEchoCancellerAdapter> echo_canceller;

std::unique_ptr<VoicePipeline> pipeline;

sc_event_fn event_fn;

void* event_context;

};

// ---------------------------------------------------------------------------

// Event type mapping

// ---------------------------------------------------------------------------

static sc_event_type_t map_event_type(EventType type) {

switch (type) {

case EventType::SessionCreated: return SC_EVENT_SESSION_CREATED;

case EventType::SpeechStarted: return SC_EVENT_SPEECH_STARTED;

case EventType::SpeechEnded: return SC_EVENT_SPEECH_ENDED;

case EventType::PartialTranscription: return SC_EVENT_PARTIAL_TRANSCRIPTION;

case EventType::TranscriptionCompleted: return SC_EVENT_TRANSCRIPTION_COMPLETED;

case EventType::ResponseCreated: return SC_EVENT_RESPONSE_CREATED;

case EventType::ResponseInterrupted: return SC_EVENT_RESPONSE_INTERRUPTED;

case EventType::ResponseAudioDelta: return SC_EVENT_RESPONSE_AUDIO_DELTA;

case EventType::ResponseDone: return SC_EVENT_RESPONSE_DONE;

case EventType::ToolCallStarted: return SC_EVENT_TOOL_CALL_STARTED;

case EventType::ToolCallCompleted: return SC_EVENT_TOOL_CALL_COMPLETED;

case EventType::Error: return SC_EVENT_ERROR;

default: return SC_EVENT_ERROR;

}

// ---------------------------------------------------------------------------

// C API implementation

// ---------------------------------------------------------------------------

extern "C" {

sc_config_t sc_config_default(void) {

sc_config_t c = {};

c.vad_onset = 0.5f;

c.vad_offset = 0.35f;

c.min_speech_duration = 0.25f;

c.min_silence_duration = 0.1f;

c.allow_interruptions = true;

c.min_interruption_duration = 1.0f;

c.interruption_recovery_timeout = 0.4f;

c.max_utterance_duration = 15.0f;

c.pre_speech_buffer_duration = 0.6f;

c.max_response_duration = 10.0f;

c.post_playback_guard = 0.3f;

c.eager_stt = true;

c.eager_stt_delay = 0.3f;

c.warmup_stt = true;

c.max_history_messages = 50;

c.max_history_tokens = 0;

c.mask_tool_results = true;

c.emit_partial_transcriptions = false;

c.partial_transcription_interval = 1.0f;

c.language = "";

c.mode = SC_MODE_ECHO;

return c;

}

sc_pipeline_t sc_pipeline_create(

sc_stt_vtable_t stt,

sc_tts_vtable_t tts,

sc_llm_vtable_t* llm,

sc_vad_vtable_t vad,

sc_config_t config,

sc_event_fn on_event,

void* event_context)

{

auto p = new sc_pipeline_s();

p->event_fn = on_event;

p->event_context = event_context;

// Create adapters

p->stt = std::make_unique<CSTTAdapter>(stt);

p->tts = std::make_unique<CTTSAdapter>(tts);

p->vad = std::make_unique<CVADAdapter>(vad);

if (llm) {

p->llm = std::make_unique<CLLMAdapter>(*llm);

}

// Convert config

AgentConfig agent_config;

agent_config.vad.onset = config.vad_onset;

agent_config.vad.offset = config.vad_offset;

agent_config.vad.min_speech_duration = config.min_speech_duration;

agent_config.vad.min_silence_duration = config.min_silence_duration;

agent_config.vad.pre_speech_buffer_duration = config.pre_speech_buffer_duration;

agent_config.allow_interruptions = config.allow_interruptions;

agent_config.min_interruption_duration = config.min_interruption_duration;

agent_config.interruption_recovery_timeout = config.interruption_recovery_timeout;

agent_config.max_utterance_duration = config.max_utterance_duration;

agent_config.max_response_duration = config.max_response_duration;

agent_config.post_playback_guard = config.post_playback_guard;

agent_config.eager_stt = config.eager_stt;

agent_config.eager_stt_delay = config.eager_stt_delay;

agent_config.warmup_stt = config.warmup_stt;

agent_config.max_history_messages = config.max_history_messages;

agent_config.max_history_tokens = config.max_history_tokens;

agent_config.mask_tool_results = config.mask_tool_results;

agent_config.emit_partial_transcriptions = config.emit_partial_transcriptions;

agent_config.partial_transcription_interval = config.partial_transcription_interval;

agent_config.language = config.language ? config.language : "";

agent_config.mode = static_cast<AgentConfig::Mode>(config.mode);

// Create pipeline

p->pipeline = std::make_unique<VoicePipeline>(

*p->stt, *p->tts, p->llm.get(), *p->vad,

agent_config,

[p](const PipelineEvent& event) {

sc_event_t e = {};

e.type = map_event_type(event.type);

e.text = event.text.c_str();

e.audio_data = event.audio_data.data();

e.audio_data_length = event.audio_data.size();

e.start_time = event.start_time;

e.end_time = event.end_time;

e.confidence = event.confidence;

e.stt_duration_ms = event.stt_duration_ms;

e.llm_duration_ms = event.llm_duration_ms;

e.tts_duration_ms = event.tts_duration_ms;

p->event_fn(&e, p->event_context);

});

// Wire token counter from LLM vtable if available

if (llm && llm->count_tokens && config.max_history_tokens > 0) {

auto count_fn = llm->count_tokens;

auto llm_ctx = llm->context;

p->pipeline->conversation_context().set_token_counter(

[count_fn, llm_ctx](const std::string& text) -> int {

return count_fn(llm_ctx, text.c_str());

});

}

return p;

}

void sc_pipeline_destroy(sc_pipeline_t pipeline) {

delete pipeline;

}

void sc_pipeline_start(sc_pipeline_t pipeline) {

if (pipeline) pipeline->pipeline->start();

}

void sc_pipeline_stop(sc_pipeline_t pipeline) {

if (pipeline) pipeline->pipeline->stop();

}

void sc_pipeline_cancel_current_turn(sc_pipeline_t pipeline) {

if (pipeline) pipeline->pipeline->cancel_current_turn();

}

void sc_pipeline_push_audio(sc_pipeline_t pipeline,

const float* samples, size_t count)

{

if (pipeline) pipeline->pipeline->push_audio(samples, count);

}

void sc_pipeline_resume_listening(sc_pipeline_t pipeline) {

if (pipeline) pipeline->pipeline->resume_listening();

}

void sc_pipeline_push_text(sc_pipeline_t pipeline, const char* text) {

if (pipeline && text) pipeline->pipeline->push_text(std::string(text));

}

sc_state_t sc_pipeline_state(sc_pipeline_t pipeline) {

if (!pipeline) return SC_STATE_IDLE;

return static_cast<sc_state_t>(pipeline->pipeline->state());

}

bool sc_pipeline_is_running(sc_pipeline_t pipeline) {

if (!pipeline) return false;

return pipeline->pipeline->is_running();

}

void sc_pipeline_set_enhancer(sc_pipeline_t pipeline,

sc_enhancer_vtable_t enhancer) {

if (!pipeline) return;

pipeline->enhancer = std::make_unique<CEnhancerAdapter>(enhancer);

pipeline->pipeline->set_enhancer(pipeline->enhancer.get());

}

void sc_pipeline_set_echo_canceller(sc_pipeline_t pipeline,

sc_echo_canceller_vtable_t aec) {

if (!pipeline) return;

pipeline->echo_canceller = std::make_unique<CEchoCancellerAdapter>(aec);

pipeline->pipeline->set_echo_canceller(pipeline->echo_canceller.get());

}

void sc_pipeline_add_tool(sc_pipeline_t pipeline, sc_tool_definition_t tool) {

if (!pipeline) return;

ToolDefinition def;

def.name = tool.name ? tool.name : "";

def.description = tool.description ? tool.description : "";

def.timeout = tool.timeout;

def.cooldown = tool.cooldown;

// Copy triggers

if (tool.triggers) {

for (const char** t = tool.triggers; *t != nullptr; t++) {

def.triggers.push_back(*t);

}

if (tool.handler) {

// Capture the C callback + context as a C++ handler

auto c_handler = tool.handler;

auto c_ctx = tool.handler_context;

def.handler = [c_handler, c_ctx](const std::string& name,

const std::string& args) -> std::string {

const char* result = c_handler(name.c_str(), args.c_str(), c_ctx);

return result ? std::string(result) : "";

};

} else if (tool.command) {

def.command = tool.command;

}

pipeline->pipeline->tool_registry().add(std::move(def));

}

int sc_pipeline_load_tools_json(sc_pipeline_t pipeline, const char* json) {

if (!pipeline || !json) return -1;

return pipeline->pipeline->tool_registry().load_json(std::string(json));

}

void sc_pipeline_clear_tools(sc_pipeline_t pipeline) {

if (!pipeline) return;

pipeline->pipeline->tool_registry().clear();

}

} // extern "C"

speech-core/src/speech_core_c.cpp at main · soniqo/speech-core · GitHub

Uh oh!

Expand file tree

Breadcrumbs

speech_core_c.cpp

Latest commit

History

Breadcrumbs

speech_core_c.cpp

File metadata and controls