Phase 12~13 전체 구현: Pro+ 피처 6종 + 음성 대화 + RAG + OS 자동화

Phase 12:
- FileTranscriptionService: ffmpeg PCM 변환 + 30초 청크 순차 STT
- MeetingSummaryService: 자막 세션 → LLM 자동 요약 + DB summaryText
- DictationTemplateService: 필드별 음성 입력 상태 머신 + 프리셋 3개

Phase 13.1:
- VoiceConversationService: STT→Ollama /api/chat→TTS 대화 루프 (10턴)
- TTSPlaybackService: Windows SAPI 문장 단위 큐 재생
- LocalLLMService.chatStream: Ollama /api/chat 스트리밍

Phase 13.2:
- RAGService: Ollama 임베딩 + SQLite 벡터 + 코사인 유사도 검색
- KnowledgeBasePage: 문서 관리 + 질문/답변 UI
- PDF 파서: zlib FlateDecode 해제 + BT/ET 텍스트 추출

Phase 13.3:
- VoiceActionService: LLM JSON 액션 플랜 생성 + 실행
- 프리셋 6개 (크롬/메모장/탐색기/볼륨), 위험 명령 차단

공통: IPC ~70채널, 에러코드 780-878, i18n 100+키
버그픽스: 라이선스 로컬 키 우선, i18n featureLabel, DOM 중첩
This commit is contained in:
Yun Chan 2026-04-05 23:52:14 +09:00
parent a31f96bbb8
commit eb83682269
38 changed files with 5678 additions and 19 deletions

View file

@ -378,6 +378,81 @@ class LocalLLMService extends EventEmitter {
return this._available
}
/**
* Ollama /api/chat .
* messages .
* yield, return.
*/
async *chatStream(
messages: Array<{ role: string; content: string }>,
options?: { model?: string; temperature?: number },
): AsyncGenerator<string, string> {
if (!this._available) {
throw new D3ROError(ErrorCode.LLMServerUnreachable, 'Ollama server not available')
}
const serverUrl = configGet('ollamaServerUrl')
const model = options?.model ?? configGet('llmModelId') ?? 'qwen3:4b'
this._abortController = new AbortController()
this._state = LLMState.Generating
try {
const response = await fetch(`${serverUrl}/api/chat`, {
method: 'POST',
headers: { 'Content-Type': 'application/json' },
body: JSON.stringify({
model,
messages,
stream: true,
options: {
temperature: options?.temperature ?? 0.7,
},
}),
signal: this._abortController.signal,
})
if (!response.ok || !response.body) {
throw new D3ROError(ErrorCode.LLMProcessingFailed, `Chat API error: ${response.status}`)
}
const reader = response.body.getReader()
const decoder = new TextDecoder()
let buffer = ''
let accumulated = ''
while (true) {
const { done, value } = await reader.read()
if (done) break
buffer += decoder.decode(value, { stream: true })
const lines = buffer.split('\n')
buffer = lines.pop() ?? ''
for (const line of lines) {
if (!line.trim()) continue
try {
const chunk = JSON.parse(line) as { message?: { content: string }; done: boolean }
if (chunk.message?.content) {
accumulated += chunk.message.content
yield chunk.message.content
}
if (chunk.done) {
return accumulated
}
} catch {
// 불완전 JSON 무시
}
}
}
return accumulated
} finally {
this._state = this._available ? LLMState.Available : LLMState.Unavailable
this._abortController = null
}
}
dispose(): void {
this._disposed = true
this.stopPolling()