auracaster/auracast-translator
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

basic working cloud model

Browse Source
This commit is contained in:
pstruebi
2025-03-06 18:03:22 +01:00
parent 1c60007f8d
commit 838e62d16e
2 changed files with 31 additions and 40 deletions
Download Patch File Download Diff File Expand all files Collapse all files

69
src/multilang_translator/main_cloud.py
View File

@@ -16,47 +16,39 @@ import voice_models.request_models
def announcement_from_german_text(
global_config: auracast_config.AuracastGlobalConfig,
translator_config: List[translator_config.TranslatorConfigDe],
config: translator_config.TranslatorConfigGroup,
text_de
):
base_lang = "deu"
audio_data_dict = {}
for i, trans in enumerate(translator_config):
if trans.big.language == base_lang:
for i, big in enumerate(config.bigs):
if big.language == base_lang:
text = text_de
else:
text = llm_translator.translate_de_to_x(
text_de,
trans.big.language,
model=trans.translator_llm,
client = trans.llm_client,
host=trans.llm_host_url,
token=trans.llm_host_token
big.language,
model=big.translator_llm,
client = big.llm_client,
host=big.llm_host_url,
token=big.llm_host_token
)
log.info('%s', text)
request_data = voice_models.request_models.SynthesizeRequest(
text=text,
target_sample_rate=global_config.auracast_sampling_rate_hz,
framework=trans.tts_system,
model=trans.tts_model,
target_sample_rate=config.auracast_sampling_rate_hz,
framework=big.tts_system,
model=big.tts_model,
return_lc3=True
)
start = time.time()
lc3_audio = voice_client.tts_client.request_synthesis(
request_data
)
# lc3_audio = text_to_speech.synthesize(
# text,
# global_config.auracast_sampling_rate_hz,
# trans.tts_system,
# trans.tts_model,
# return_lc3=True
# )
audio_data_dict[trans.big.language] = lc3_audio.decode('latin-1') # TODO: should be .hex in the future
log.info('Voice synth took %s', time.time() - start)
audio_data_dict[big.language] = lc3_audio.decode('latin-1') # TODO: should be .hex in the future
multicast_client.send_audio(
audio_data_dict
@@ -69,27 +61,26 @@ if __name__ == '__main__':
format='%(module)s.py:%(lineno)d %(levelname)s: %(message)s'
)
global_conf = auracast_config.AuracastGlobalConfig()
#global_conf.transport='serial:/dev/serial/by-id/usb-SEGGER_J-Link_001057705357-if02,1000000,rtscts' # transport for nrf54l15dk
global_conf.transport='serial:/dev/serial/by-id/usb-ZEPHYR_Zephyr_HCI_UART_sample_81BD14B8D71B5662-if00,115200,rtscts' #nrf52dongle hci_uart usb cdc
translator_conf = [
translator_config.TranslatorConfigDe(),
translator_config.TranslatorConfigEn(),
translator_config.TranslatorConfigFr(),
#auracast_config.broadcast_es,
#auracast_config.broadcast_it,
config = translator_config.TranslatorConfigGroup(
bigs=[
translator_config.TranslatorConfigDe(),
translator_config.TranslatorConfigEn(),
translator_config.TranslatorConfigFr(),
]
for conf in translator_conf:
conf.big.loop = False
)
config.transport='serial:/dev/serial/by-id/usb-ZEPHYR_Zephyr_HCI_UART_sample_81BD14B8D71B5662-if00,115200,rtscts' #nrf52dongle hci_uart usb cdc
for conf in config.bigs:
conf.loop = False
conf.llm_client = 'openwebui' # comment out for local llm
conf.llm_host_url = 'https://ollama.pstruebi.xyz'
conf.llm_host_token = 'sk-17124cb84df14cc6ab2d9e17d0724d13'
multicast_client.request_init(
config
)
multicast_client.initialize()
announcement_from_german_text(global_conf, translator_conf, 'Hello')
announcement_from_german_text(config, 'Hello')
# TODO: make everything async

2
src/multilang_translator/translator_config.py
View File

@@ -5,7 +5,7 @@ from auracast import auracast_config
VENV_DIR = os.path.join(os.path.dirname(__file__), './../../venv')
class TranslatorLangConfig(BaseModel):
class TranslatorLangConfig(auracast_config.AuracastBigConfig):
translator_llm: str = 'llama3.2:3b-instruct-q4_0'
llm_client: str = 'ollama'
llm_host_url: str | None = 'http://localhost:11434'