feat: add ASR filtering options

custom_components/tencentcloud_asr/__init__.py

@@ -13,6 +13,10 @@ _LOGGER = logging.getLogger(__name__)
 SecretIdKey = "secretid"
 SecretKeyKey = "secretkey"
 ModelKey = "model"
+FilterDirtyKey = "filter_dirty"
+FilterModalKey = "filter_modal"
+FilterPuncKey = "filter_punc"
+ConvertNumModeKey = "convert_num_mode"
 
 
 async def async_setup_entry(hass: HomeAssistant, config_entry: ConfigEntry):

custom_components/tencentcloud_asr/config_flow.py

@@ -3,11 +3,24 @@ import voluptuous as vol
 from typing import Any, Dict
 from homeassistant.core import callback
 from homeassistant.config_entries import ConfigFlow, ConfigFlowResult, OptionsFlow, ConfigEntry
-from homeassistant.helpers.selector import SelectSelector, SelectSelectorConfig, SelectOptionDict
+from homeassistant.helpers.selector import (
+    BooleanSelector,
+    SelectOptionDict,
+    SelectSelector,
+    SelectSelectorConfig,
+)
 
 from .tencentcloud_api import ModuleSupportLanguage, DefaultModel
 from .tencentcloud_api import TencentCloudAsrAPi
-from . import SecretIdKey, SecretKeyKey, ModelKey
+from . import (
+    ConvertNumModeKey,
+    FilterDirtyKey,
+    FilterModalKey,
+    FilterPuncKey,
+    ModelKey,
+    SecretIdKey,
+    SecretKeyKey,
+)
 from . import DOMAIN
 
 _LOGGER = logging.getLogger(__name__)
@@ -22,6 +35,17 @@ modelSelector = SelectSelector(
     )
 )
 
+filterDirtySelector = SelectSelector(
+    SelectSelectorConfig(
+        options=[
+            SelectOptionDict(value="0", label="0"),
+            SelectOptionDict(value="1", label="1"),
+            SelectOptionDict(value="2", label="2"),
+        ],
+        translation_key="filter_dirty_descriptions",
+    )
+)
+
 
 class ConfigFlowHandler(ConfigFlow, domain=DOMAIN):
     async def async_step_user(self, user_input: dict[str, Any] | None = None) -> ConfigFlowResult:
@@ -43,6 +67,10 @@ class ConfigFlowHandler(ConfigFlow, domain=DOMAIN):
                 vol.Required(SecretIdKey): str,
                 vol.Required(SecretKeyKey): str,
                 vol.Required(ModelKey, default=DefaultModel): modelSelector,
+                vol.Required(FilterDirtyKey, default="0"): filterDirtySelector,
+                vol.Required(FilterModalKey, default=False): BooleanSelector(),
+                vol.Required(FilterPuncKey, default=False): BooleanSelector(),
+                vol.Required(ConvertNumModeKey, default=True): BooleanSelector(),
             })
 
         return self.async_show_form(step_id="user", data_schema=config_schema, errors=errors)
@@ -68,5 +96,21 @@ class TencentCloudAsrOptionFlow(OptionsFlow):
         model = user_input.get(ModelKey, DefaultModel)
         config_schema = vol.Schema({
             vol.Optional(ModelKey, default=model): modelSelector,
+            vol.Required(
+                FilterDirtyKey,
+                default=user_input.get(FilterDirtyKey, "0"),
+            ): filterDirtySelector,
+            vol.Required(
+                FilterModalKey,
+                default=user_input.get(FilterModalKey, False),
+            ): BooleanSelector(),
+            vol.Required(
+                FilterPuncKey,
+                default=user_input.get(FilterPuncKey, False),
+            ): BooleanSelector(),
+            vol.Required(
+                ConvertNumModeKey,
+                default=user_input.get(ConvertNumModeKey, True),
+            ): BooleanSelector(),
         })
         return self.async_show_form(step_id="user", data_schema=config_schema, errors={})

custom_components/tencentcloud_asr/stt.py

@@ -10,7 +10,15 @@ from homeassistant.helpers.entity_platform import AddEntitiesCallback
 from .tencentcloud_api import TencentCloudAsrAPi
 from .tencentcloud_api import ModuleSupportLanguage
 from .tencentcloud_api import DefaultModel
-from . import SecretIdKey, SecretKeyKey, ModelKey
+from . import (
+    ConvertNumModeKey,
+    FilterDirtyKey,
+    FilterModalKey,
+    FilterPuncKey,
+    ModelKey,
+    SecretIdKey,
+    SecretKeyKey,
+)
 
 _LOGGER = logging.getLogger(__name__)
 
@@ -32,6 +40,7 @@ class ASRSTT(stt.SpeechToTextEntity):
         self._attr_unique_id = f"tencentcloud_asr_stt"
         self._attr_name = f"TencentCloud Asr STT"
         self.hass = hass
+        self.config_entry = config_entry
 
     @property
     def supported_languages(self) -> list[str]:
@@ -71,9 +80,25 @@ class ASRSTT(stt.SpeechToTextEntity):
             audio += chunk
 
         data = base64.b64encode(audio).decode("utf8", "ignore").strip()
-        model: str = self.hass.data.get(ModelKey, DefaultModel)
+        config = {**self.config_entry.data, **self.config_entry.options}
-        _LOGGER.debug(f"process_audio_stream transcribe: {data} {model}")
+        model: str = config.get(ModelKey, DefaultModel)
-        ret, result = await self.hass.async_add_executor_job(partial(self.tencentCloudApi.SentenceRecognition, model, data))
+        _LOGGER.debug(
+            "process_audio_stream transcribe: audio_bytes=%s model=%s",
+            len(audio),
+            model,
+        )
+        ret, result = await self.hass.async_add_executor_job(
+            partial(
+                self.tencentCloudApi.SentenceRecognition,
+                model,
+                data,
+                len(audio),
+                config.get(FilterDirtyKey, "0"),
+                config.get(FilterModalKey, False),
+                config.get(FilterPuncKey, False),
+                config.get(ConvertNumModeKey, True),
+            )
+        )
         if not ret:
             return stt.SpeechResult(result, stt.SpeechResultState.ERROR)
         if not result:

custom_components/tencentcloud_asr/tencentcloud_api.py

@@ -43,7 +43,16 @@ class TencentCloudAsrAPi(object):
         region = ""
         self.asrClient = AsrClient(cred, region)
 
-    def SentenceRecognition(self, engSerViceType, data: str) -> (tuple[bool, str | None]):
+    def SentenceRecognition(
+        self,
+        engSerViceType,
+        data: str,
+        data_len: int,
+        filter_dirty: str,
+        filter_modal: bool,
+        filter_punc: bool,
+        convert_num_mode: bool,
+    ) -> tuple[bool, str | None]:
         req = SentenceRecognitionRequest()
         req.EngSerViceType = engSerViceType
         req.SourceType = SentenceRecognitionSourceTypePost
@@ -52,7 +61,11 @@ class TencentCloudAsrAPi(object):
         req.SubServiceType = 2
         req.UsrAudioKey = ""
         req.Data = data
-        req.DataLen = len(data)
+        req.DataLen = data_len
+        req.FilterDirty = int(filter_dirty)
+        req.FilterModal = int(filter_modal)
+        req.FilterPunc = int(filter_punc)
+        req.ConvertNumMode = int(convert_num_mode)
         try:
             res: SentenceRecognitionResponse = self.asrClient.SentenceRecognition(req)
             return True, res.Result

custom_components/tencentcloud_asr/translations/en.json

@@ -7,7 +7,11 @@
                 "data": {
                     "secretid": "tencentcloud asr secretid",
                     "secretkey": "tencnetcloud asr secretkey",
-                    "model": "model"
+                    "model": "model",
+                    "filter_dirty": "Dirty word handling",
+                    "filter_modal": "Filter filler words such as ‘um’ and ‘uh’",
+                    "filter_punc": "Filter sentence-ending punctuation",
+                    "convert_num_mode": "Smart Arabic numeral conversion"
                 },
                 "description": "The integration is based on [Tencent Cloud API](https://cloud.tencent.com/document/product/1093/35646)."
             }
@@ -17,12 +21,23 @@
         "step": {
             "user": {
                 "data": {
-                    "model": "model"
+                    "model": "model",
+                    "filter_dirty": "Dirty word handling",
+                    "filter_modal": "Filter filler words such as ‘um’ and ‘uh’",
+                    "filter_punc": "Filter sentence-ending punctuation",
+                    "convert_num_mode": "Smart Arabic numeral conversion"
                 }
             }
         }
     },
     "selector": {
+        "filter_dirty_descriptions": {
+            "options": {
+                "0": "Do not filter dirty words",
+                "1": "Filter dirty words",
+                "2": "Replace dirty words with *"
+            }
+        },
         "model_descriptions": {
             "options": {
                 "8k_zh": "8k_zh - Chinese telephone audio",

custom_components/tencentcloud_asr/translations/zh-Hans.json

@@ -7,7 +7,11 @@
                 "data": {
                     "secretid": "请输入腾讯云ASR服务secretid",
                     "secretkey": "请输入腾讯云ASR服务secretkey",
-                    "model": "模型"
+                    "model": "模型",
+                    "filter_dirty": "脏词处理方式",
+                    "filter_modal": "过滤“嗯、啊”等语气词",
+                    "filter_punc": "过滤句末标点",
+                    "convert_num_mode": "阿拉伯数字智能转换"
                 },
                 "description": "此集成基于腾讯云API, 具体见[腾讯云文档](https://cloud.tencent.com/document/product/1093/35646)"
             }
@@ -17,12 +21,23 @@
         "step": {
             "user": {
                 "data": {
-                    "model": "模型"
+                    "model": "模型",
+                    "filter_dirty": "脏词处理方式",
+                    "filter_modal": "过滤“嗯、啊”等语气词",
+                    "filter_punc": "过滤句末标点",
+                    "convert_num_mode": "阿拉伯数字智能转换"
                 }
             }
         }
     },
     "selector": {
+        "filter_dirty_descriptions": {
+            "options": {
+                "0": "不过滤脏词",
+                "1": "过滤脏词",
+                "2": "将脏词替换为 *"
+            }
+        },
         "model_descriptions": {
             "options": {
                 "8k_zh": "8k_zh - 中文电话通用",