Agent Zero AI framework
Alpine fix version, STT fixes
frdelc99b1a47d4f2
diff --git a/python/helpers/settings.py b/python/helpers/settings.pyindex 867969ab7..af0627367 100644--- a/python/helpers/settings.py+++ b/python/helpers/settings.py@@ -4,6 +4,7 @@ import refrom typing import Any, Literal, Optional, TypedDictimport models+from python.helpers import whisperfrom . import files, dotenvfrom models import get_model, ModelProvider, ModelTypefrom langchain_core.language_models.chat_models import BaseChatModel@@ -44,7 +45,6 @@ class Settings(TypedDict):stt_silence_threshold: floatstt_silence_duration: intstt_waiting_timeout: int- stt_min_speech_duration: intclass PartialSettings(Settings, total=False):@@ -474,14 +474,6 @@ def convert_out(settings: Settings) -> SettingsOutput:"value": settings["stt_waiting_timeout"]})- stt_fields.append({- "id": "stt_min_speech_duration",- "title": "Prefix padding (ms)",- "description": "Minimum duration of audio to be included in the stream before speech was recognized.",- "type": "input",- "value": settings["stt_min_speech_duration"]- })-stt_section: SettingsSection = {"title": "Speech to Text","description": "Voice transcription preferences and server turn detection settings.",@@ -593,12 +585,6 @@ def get_embedding_model(settings: Settings | None = None) -> Embeddings:**settings["embed_model_kwargs"],)-def get_speech_settings(settings: Settings | None = None) -> dict:- if not settings:- settings = get_settings()- return settings["speech_to_text"]--def _read_settings_file() -> Settings | None:if os.path.exists(SETTINGS_FILE):content = files.read_file(SETTINGS_FILE)@@ -653,12 +639,11 @@ def _get_default_settings() -> Settings:agent_knowledge_subdir="custom",rfc_url="http://localhost:55080",rfc_password="",- stt_model_size="tiny",+ stt_model_size="base",stt_language="en",- stt_silence_threshold=0.15,+ stt_silence_threshold=0.3,stt_silence_duration=1000,stt_waiting_timeout=2000,- stt_min_speech_duration=500)@@ -676,6 +661,8 @@ def _apply_settings():agent.config = ctx.configagent = agent.get_data(agent.DATA_NAME_SUBORDINATE)+ # reload whisper model if necessary+ whisper.preload()def _env_to_dict(data: str):env_dict = {}diff --git a/python/helpers/whisper.py b/python/helpers/whisper.pyindex 0b793aeb5..97bb89b99 100644--- a/python/helpers/whisper.py+++ b/python/helpers/whisper.py@@ -3,16 +3,20 @@ import base64import warningsimport whisperimport tempfile-from python.helpers import runtime, rfc+from python.helpers import runtime, rfc, settings# suppress FutureWarning from torch.loadwarnings.filterwarnings('ignore', category=FutureWarning)model = None+model_name = ""def preload():- global model- model = whisper.load_model("base")+ global model, model_name+ set = settings.get_settings()+ if not model or model_name != set["stt_model_size"]:+ model = whisper.load_model(set["stt_model_size"])+ model_name = set["stt_model_size"]return modelasync def transcribe(audio_bytes_b64: str):diff --git a/webui/index.html b/webui/index.htmlindex 86554193b..0ad481e78 100644--- a/webui/index.html+++ b/webui/index.html@@ -19,8 +19,8 @@}</script>- <script defer src="https://cdn.jsdelivr.net/npm/@alpinejs/[email protected]/dist/cdn.min.js"></script>- <script defer src="https://cdn.jsdelivr.net/npm/[email protected]/dist/cdn.min.js"></script>+ <script defer src="https://cdn.jsdelivr.net/npm/@alpinejs/[email protected]/dist/cdn.min.js"></script>+ <script defer src="https://cdn.jsdelivr.net/npm/[email protected]/dist/cdn.min.js"></script><script src="https://cdn.jsdelivr.net/npm/[email protected]/src-noconflict/ace.js"></script><link href="https://cdn.jsdelivr.net/npm/[email protected]/css/ace.min.css" rel="stylesheet">diff --git a/webui/js/settings.js b/webui/js/settings.jsindex 8be6fd542..e5eb65149 100644--- a/webui/js/settings.js+++ b/webui/js/settings.js@@ -44,7 +44,7 @@ const settingsModalProxy = {const modalEl = document.getElementById('settingsModal');const modalAD = Alpine.$data(modalEl);resp = await window.sendJsonData("/settings_set", modalAD.settings);-+ document.dispatchEvent(new CustomEvent('settings-updated', { detail: resp.settings }));this.resolvePromise({status: 'saved',data: resp.settingsdiff --git a/webui/js/speech.js b/webui/js/speech.jsindex d7118b131..602838e3a 100644--- a/webui/js/speech.js+++ b/webui/js/speech.js@@ -14,6 +14,38 @@ const Status = {PROCESSING: 'processing'};+const micSettings = {+ stt_model_size: 'tiny',+ stt_language: 'en',+ stt_silence_threshold: 0.05,+ stt_silence_duration: 1000,+ stt_waiting_timeout: 2000,+};+window.micSettings = micSettings+loadMicSettings()++function densify(x) {+ return Math.exp(-5 * (1 - x));+}++async function loadMicSettings() {+ try {+ const response = await fetch('/settings_get');+ const data = await response.json();+ const sttSettings = data.settings.sections.find(s => s.title === 'Speech to Text');++ if (sttSettings) {+ // Update options from server settings+ sttSettings.fields.forEach(field => {+ const key = field.id //.split('.')[1]; // speech_to_text.model_size -> model_size+ micSettings[key] = field.value;+ });+ }+ } catch (error) {+ console.error('Failed to load speech settings:', error);+ }+}+class MicrophoneInput {constructor(updateCallback, options = {}) {this.mediaRecorder = null;@@ -34,37 +66,6 @@ class MicrophoneInput {this.silenceStartTime = null;this.hasStartedRecording = false;this.analysisFrame = null;-- // Initialize with defaults- this.options = {- modelSize: 'tiny',- language: 'en',- silenceThreshold: 0.15,- silenceDuration: 1000,- waitingTimeout: 2000,- minSpeechDuration: 500- };-- // Fetch settings from server- this.loadSettings();- }-- async loadSettings() {- try {- const response = await fetch('/settings_get');- const data = await response.json();- const sttSettings = data.settings.sections.find(s => s.title === 'Speech to Text');-- if (sttSettings) {- // Update options from server settings- sttSettings.fields.forEach(field => {- const key = field.id.split('.')[1]; // speech_to_text.model_size -> model_size- this.options[key] = field.value;- });- }- } catch (error) {- console.error('Failed to load speech settings:', error);- }}get status() {@@ -148,7 +149,7 @@ class MicrophoneInput {if (this.status === Status.WAITING) {this.status = Status.PROCESSING;}- }, this.options.waitingTimeout);+ }, micSettings.stt_waiting_timeout);}handleProcessingState() {@@ -228,7 +229,7 @@ class MicrophoneInput {const now = Date.now();// Update status based on audio level- if (rms > this.options.silenceThreshold) {+ if (rms > densify(micSettings.stt_silence_threshold)) {this.lastAudioTime = now;this.silenceStartTime = null;@@ -242,7 +243,7 @@ class MicrophoneInput {}const silenceDuration = now - this.silenceStartTime;- if (silenceDuration >= this.options.silenceDuration) {+ if (silenceDuration >= micSettings.stt_silence_duration) {this.status = Status.WAITING;}}@@ -326,7 +327,7 @@ class MicrophoneInput {// Initialize and handle click eventsasync function initializeMicrophoneInput() {- microphoneInput = new MicrophoneInput(+ window.microphoneInput = microphoneInput = new MicrophoneInput(async (text, isFinal) => {if (isFinal) {updateChatInput(text);@@ -456,8 +457,4 @@ export const speech = new Speech();window.speech = speech// Add event listener for settings changes-document.addEventListener('settings-updated', async () => {- if (microphoneInput) {- await microphoneInput.loadSettings();- }-});\ No newline at end of file+document.addEventListener('settings-updated', loadMicSettings);\ No newline at end of file