(user_input)

(text: str, timeout=2.0)

(self, model_path: str, codec_path: str)

(llasa_1b, codec_path)

( age, gender, pitch, pitch_var, volume, speed, emo )

(speech_tokens_str)

(speech_tokens_str)

(wav)

(text: str)

( model, codec_model, device, tokenizer, refined_text, instruct_text, control_tag

()

(text: str, timeout=2.0)

(self, db_path, collection_name)

(self, model_path: str, codec_path: str, devices: List[str])

(self)

(tpl_name)

(tpl_name)

(error)

(wav_data)

(qwen3emb, text: str)

(speech_ids)

(refined_text: str, instruct_text: str, ref_wav: str = None, user_instruct_tag: bool = False, cs2_if_rag_text:

( refined_text, instruct_text, if_rag_text, age, gender, pitch, pitch_var, vol

(user_input)

(user_input, use_normalization)

(req: RagQuery)

(self)

(user_provide)

(user_provide)

(sample_rate=48000, bit_depth=16, channels=1)

( worker_id: int, device: str, model_path: str, codec_path: str, task_queue: multiproces