feat: add ASR filtering options

.github/workflows/release.yaml

@@ -1,25 +1,32 @@
 name: Release
 
 on:
-  release:
+  push:
-    types: [published]
+    tags:
+      - "v*"
+
+permissions:
+  contents: write
 
 jobs:
   release-zip:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v2
+      - name: Checkout code
+        uses: actions/checkout@v4
 
       - name: ZIP Component Dir
         run: |
           cd ${{ github.workspace }}/custom_components/tencentcloud_asr
-          zip -r tencentcloud_asr.zip ./
+          zip -r ${{ github.workspace }}/tencentcloud_asr.zip ./
 
-      - name: Upload zip to release
+      - name: Create Release
-        uses: svenstaro/upload-release-action@v2
+        uses: softprops/action-gh-release@v2
         with:
-          repo_token: ${{ secrets.GITHUB_TOKEN }}
+          tag_name: ${{ github.ref_name }}
-          file: ${{ github.workspace }}/custom_components/tencentcloud_asr/tencentcloud_asr.zip
+          name: Release ${{ github.ref_name }}
-          asset_name: tencentcloud_asr.zip
+          draft: false
-          tag: ${{ github.ref }}
+          prerelease: false
-          overwrite: true
+          files: tencentcloud_asr.zip
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

README.md

@@ -2,6 +2,10 @@
 
 腾讯云语音识别服务 homeassistant stt集成
 
+[![GitHub release][releases-shield]][releases]
+[![Github downloads][downloads-shield]][releases]
+[![HACS][hacs-shield]][hacs]
+
 ## 安装
 
 - 腾讯云上开通[语音识别服务](https://cloud.tencent.com/product/asr)，获取SecretId 和 SecretKey
@@ -18,12 +22,36 @@
 
 | 模型 | 说明 |
 |------|------|
+|8k_zh|中文电话通用|
+|8k_en|英文电话通用|
 |16k_zh| 中文通用|
 |16k_zh-PY|中英粤|
+|16k_zh_medical|中文医疗|
 |16k_en|英语|
 |16k_yue|粤语|
+|16k_ja|日语|
+|16k_ko|韩语|
+|16k_vi|越南语|
+|16k_ms|马来语|
+|16k_id|印度尼西亚语|
+|16k_fil|菲律宾语|
+|16k_th|泰语|
+|16k_pt|葡萄牙语|
+|16k_tr|土耳其语|
+|16k_ar|阿拉伯语|
+|16k_es|西班牙语|
+|16k_hi|印地语|
+|16k_fr|法语|
+|16k_de|德语|
 |16k_zh_dialect|多方言，支持23种方言（上海话、四川话、武汉话、贵阳话、昆明话、西安话、郑州话、太原话、兰州话、银川话、西宁话、南京话、合肥话、南昌话、长沙话、苏州话、杭州话、济南话、天津话、石家庄话、黑龙江话、吉林话、辽宁话）|
 
 - 更多模型详细说明，请参考[腾讯云文档](https://cloud.tencent.com/document/product/1093/35646)
 - 默认模型: `16k_zh`
 
+
+
+[releases-shield]: https://img.shields.io/github/v/release/chliny/hass-tencentcloud-asr?style=flat-square
+[downloads-shield]: https://img.shields.io/github/downloads/chliny/hass-tencentcloud-asr/total?style=flat-square
+[releases]: https://github.com/chliny/hass-tencentcloud-asr/releases
+[hacs-shield]: https://img.shields.io/badge/HACS-Default-41BDF5.svg?style=flat-square
+[hacs]: https://github.com/hacs/integration

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/manifest.json

@@ -16,5 +16,5 @@
     "loggers": [
         "tencentcloud_asr"
     ],
-    "version": "0.0.1"
+    "version": "0.1.1"
 }

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]:
@@ -52,6 +61,9 @@ class ASRSTT(stt.SpeechToTextEntity):
 
     @property
     def supported_sample_rates(self) -> list[stt.AudioSampleRates]:
+        model: str = self.hass.data.get(ModelKey, DefaultModel)
+        if model.startswith("8k_"):
+            return [stt.AudioSampleRates.SAMPLERATE_8000]
         return [stt.AudioSampleRates.SAMPLERATE_16000]
 
     @property
@@ -68,9 +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

@@ -11,10 +11,27 @@ _LOGGER = logging.getLogger(__name__)
 SentenceRecognitionSourceTypePost = 1
 SentenceRecognitionSourceTypeURL = 0
 ModuleSupportLanguage = {
+    "8k_zh": ["zh-cn"],
+    "8k_en": ["en"],
     "16k_zh": ["zh-cn"],
     "16k_zh-PY": ["zh-cn", "en", "zh-hk"],
+    "16k_zh_medical": ["zh-cn"],
     "16k_en": ["en"],
     "16k_yue": ["zh-hk"],
+    "16k_ja": ["ja"],
+    "16k_ko": ["ko"],
+    "16k_vi": ["vi"],
+    "16k_ms": ["ms"],
+    "16k_id": ["id"],
+    "16k_fil": ["fil"],
+    "16k_th": ["th"],
+    "16k_pt": ["pt"],
+    "16k_tr": ["tr"],
+    "16k_ar": ["ar"],
+    "16k_es": ["es"],
+    "16k_hi": ["hi"],
+    "16k_fr": ["fr"],
+    "16k_de": ["de"],
     "16k_zh_dialect": ["zh"],
 }
 DefaultModel = "16k_zh"
@@ -26,7 +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
@@ -35,7 +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)."
             }
@@ -17,18 +21,46 @@
         "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": {
-                "16k_zh": "16k_zh Chinese",
+                "8k_zh": "8k_zh - Chinese telephone audio",
+                "8k_en": "8k_en - English telephone audio",
+                "16k_zh": "16k_zh - Chinese",
                 "16k_zh-PY": "16k_zh-PY - Chinese, English, and Cantonese",
+                "16k_zh_medical": "16k_zh_medical - Medical Chinese",
                 "16k_en": "16k_en - English",
                 "16k_yue": "16k_yue - Cantonese",
+                "16k_ja": "16k_ja - Japanese",
+                "16k_ko": "16k_ko - Korean",
+                "16k_vi": "16k_vi - Vietnamese",
+                "16k_ms": "16k_ms - Malay",
+                "16k_id": "16k_id - Indonesian",
+                "16k_fil": "16k_fil - Filipino",
+                "16k_th": "16k_th - Thai",
+                "16k_pt": "16k_pt - Portuguese",
+                "16k_tr": "16k_tr - Turkish",
+                "16k_ar": "16k_ar - Arabic",
+                "16k_es": "16k_es - Spanish",
+                "16k_hi": "16k_hi - Hindi",
+                "16k_fr": "16k_fr - French",
+                "16k_de": "16k_de - German",
                 "16k_zh_dialect": "16k_zh_dialect - Chinese dialects"
             }
         }

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)"
             }
@@ -17,18 +21,46 @@
         "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",
+                "8k_en": "8k_en - 英文电话通用",
                 "16k_zh": "16k_zh - 中文通用",
                 "16k_zh-PY": "16k_zh-PY - 中英粤",
+                "16k_zh_medical": "16k_zh_medical - 中文医疗",
                 "16k_en": "16k_en - 英语",
                 "16k_yue": "16k_yue - 粤语",
+                "16k_ja": "16k_ja - 日语",
+                "16k_ko": "16k_ko - 韩语",
+                "16k_vi": "16k_vi - 越南语",
+                "16k_ms": "16k_ms - 马来语",
+                "16k_id": "16k_id - 印度尼西亚语",
+                "16k_fil": "16k_fil - 菲律宾语",
+                "16k_th": "16k_th - 泰语",
+                "16k_pt": "16k_pt - 葡萄牙语",
+                "16k_tr": "16k_tr - 土耳其语",
+                "16k_ar": "16k_ar - 阿拉伯语",
+                "16k_es": "16k_es - 西班牙语",
+                "16k_hi": "16k_hi - 印地语",
+                "16k_fr": "16k_fr - 法语",
+                "16k_de": "16k_de - 德语",
                 "16k_zh_dialect": "16k_zh_dialect - 中文多方言"
             }
         }