@fugood/bricks-project 2.24.0-beta.0 → 2.24.0-beta.2

package/compile/action-name-map.ts

@@ -601,6 +601,14 @@ export const templateActionNameMap = {
       variables: 'GENERATOR_MCP_VARIABLES',
     },
   },
+  GENERATOR_TRADITIONAL_VAD: {
+    GENERATOR_TRADITIONAL_VAD_DETECT_FILE: {
+      fileUrl: 'GENERATOR_TRADITIONAL_VAD_FILE_URL',
+    },
+    GENERATOR_TRADITIONAL_VAD_DETECT_DATA: {
+      data: 'GENERATOR_TRADITIONAL_VAD_DATA',
+    },
+  },
   GENERATOR_TTS: {
     GENERATOR_TTS_GENERATE: {
       text: 'GENERATOR_TTS_TEXT',
@@ -621,6 +629,24 @@ export const templateActionNameMap = {
       audioUri: 'GENERATOR_ONNX_STT_AUDIO_URI',
     },
   },
+  GENERATOR_ONNX_VAD: {
+    GENERATOR_ONNX_VAD_DETECT_FILE: {
+      fileUrl: 'GENERATOR_ONNX_VAD_FILE_URL',
+      threshold: 'GENERATOR_ONNX_VAD_THRESHOLD',
+      minSpeechDurationMs: 'GENERATOR_ONNX_VAD_MIN_SPEECH_DURATION_MS',
+      minSilenceDurationMs: 'GENERATOR_ONNX_VAD_MIN_SILENCE_DURATION_MS',
+      maxSpeechDurationS: 'GENERATOR_ONNX_VAD_MAX_SPEECH_DURATION_S',
+      speechPadMs: 'GENERATOR_ONNX_VAD_SPEECH_PAD_MS',
+    },
+    GENERATOR_ONNX_VAD_DETECT_DATA: {
+      data: 'GENERATOR_ONNX_VAD_DATA',
+      threshold: 'GENERATOR_ONNX_VAD_THRESHOLD',
+      minSpeechDurationMs: 'GENERATOR_ONNX_VAD_MIN_SPEECH_DURATION_MS',
+      minSilenceDurationMs: 'GENERATOR_ONNX_VAD_MIN_SILENCE_DURATION_MS',
+      maxSpeechDurationS: 'GENERATOR_ONNX_VAD_MAX_SPEECH_DURATION_S',
+      speechPadMs: 'GENERATOR_ONNX_VAD_SPEECH_PAD_MS',
+    },
+  },
   GENERATOR_SPEECH_INFERENCE: {
     GENERATOR_SPEECH_INFERENCE_TRANSCRIBE_FILE: {
       fileUrl: 'GENERATOR_SPEECH_INFERENCE_FILE_URL',
@@ -636,20 +662,6 @@ export const templateActionNameMap = {
       language: 'GENERATOR_SPEECH_INFERENCE_LANGUAGE',
       translate: 'GENERATOR_SPEECH_INFERENCE_TRANSLATE',
     },
-    GENERATOR_SPEECH_INFERENCE_TRANSCRIBE_REALTIME: {
-      prompt: 'GENERATOR_SPEECH_INFERENCE_PROMPT',
-      beamSize: 'GENERATOR_SPEECH_INFERENCE_BEAM_SIZE',
-      language: 'GENERATOR_SPEECH_INFERENCE_LANGUAGE',
-      translate: 'GENERATOR_SPEECH_INFERENCE_TRANSLATE',
-      realtimeAudioSeconds: 'GENERATOR_SPEECH_INFERENCE_REALTIME_AUDIO_SECONDS',
-      realtimeAudioSliceSeconds: 'GENERATOR_SPEECH_INFERENCE_REALTIME_AUDIO_SLICE_SECONDS',
-      realtimeAudioMinSeconds: 'GENERATOR_SPEECH_INFERENCE_REALTIME_AUDIO_MIN_SECONDS',
-      realtimeSaveAudio: 'GENERATOR_SPEECH_INFERENCE_REALTIME_SAVE_AUDIO',
-      realtimeVadEnabled: 'GENERATOR_SPEECH_INFERENCE_REALTIME_VAD_ENABLED',
-      realtimeVadMs: 'GENERATOR_SPEECH_INFERENCE_REALTIME_VAD_MS',
-      realtimeVadThold: 'GENERATOR_SPEECH_INFERENCE_REALTIME_VAD_THOLD',
-      realtimeVadFreqThold: 'GENERATOR_SPEECH_INFERENCE_REALTIME_VAD_FREQ_THOLD',
-    },
   },
   GENERATOR_VAD_INFERENCE: {
     GENERATOR_VAD_INFERENCE_DETECT_FILE: {

package/compile/index.ts

@@ -421,8 +421,9 @@ const compileAutomation = (automationMap: AutomationMap) =>
 
 export const compile = async (app: Application) => {
   await new Promise((resolve) => setImmediate(resolve, 0))
+  const timestamp = Date.now()
   const config = {
-    title: app.name,
+    title: `${app.name || 'Unknown'}(${timestamp})`,
     subspace_map: app.subspaces.reduce((subspaceMap, subspace) => {
       subspaceMap[subspace.id] = {
         title: subspace.title,
@@ -969,6 +970,7 @@ export const compile = async (app: Application) => {
     automation_map: app.automationMap
       ? compileAutomation(app.automationMap)
       : app.metadata?.TEMP_automation_map || {},
+    update_timestamp: timestamp,
   }
   return config
 }

package/package.json

@@ -1,13 +1,14 @@
 {
   "name": "@fugood/bricks-project",
-  "version": "2.24.0-beta.0",
+  "version": "2.24.0-beta.2",
   "main": "index.ts",
   "scripts": {
     "build": "bun scripts/build.js"
   },
   "dependencies": {
-    "@fugood/bricks-cli": "^2.23.0",
+    "@fugood/bricks-cli": "^2.24.0-beta.2",
     "@huggingface/gguf": "^0.3.2",
+    "@iarna/toml": "^3.0.0",
     "@modelcontextprotocol/sdk": "^1.15.0",
     "@toon-format/toon": "^2.1.0",
     "@types/escodegen": "^0.0.10",
@@ -18,5 +19,5 @@
     "lodash": "^4.17.4",
     "uuid": "^8.3.1"
   },
-  "gitHead": "27bd4a30a4dd412ab73eab13c49265f73fd76e6c"
+  "gitHead": "f63649f1cb9ae4c3d0c4ac687152bfc1be6528bd"
 }

package/tools/mcp-server.ts

@@ -1,7 +1,7 @@
 import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js'
 import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js'
 import { z } from 'zod'
-import { $ } from 'bun'
+import { $, JSON5 } from 'bun'
 import * as TOON from '@toon-format/toon'
 import Fuse from 'fuse.js'
 import { gguf } from '@huggingface/gguf'
@@ -729,7 +729,7 @@ server.tool(
 
         // Return JSON for huggingface_select to allow direct parsing by consumers
         return {
-          content: [{ type: 'text', text: JSON.stringify(result, null, 2) }],
+          content: [{ type: 'text', text: JSON5.stringify(result, null, 2) }],
         }
       }
 
@@ -842,7 +842,7 @@ server.tool(
 
         // Return JSON for huggingface_select to allow direct parsing by consumers
         return {
-          content: [{ type: 'text', text: JSON.stringify(result, null, 2) }],
+          content: [{ type: 'text', text: JSON5.stringify(result, null, 2) }],
         }
       } else {
         const result = {
@@ -875,7 +875,12 @@ server.tool(
 
         // Return JSON for huggingface_select to allow direct parsing by consumers
         return {
-          content: [{ type: 'text', text: JSON.stringify(result, null, 2) }],
+          content: [
+            {
+              type: 'text',
+              text: `${JSON5.stringify(result, null, 2)} // Please use _hfRepoInfo undocumented field with @ts-ignore`,
+            },
+          ],
         }
       }
     } catch (err: any) {

package/tools/postinstall.ts

@@ -1,5 +1,6 @@
 import { $ } from 'bun'
 import { stat, readFile, writeFile, readdir } from 'fs/promises'
+import TOML from '@iarna/toml'
 
 const cwd = process.cwd()
 
@@ -63,9 +64,7 @@ if (hasClaudeCode || hasAgentsMd) {
   await handleMcpConfigOverride(mcpConfigPath)
 }
 
-if (hasClaudeCode) {
-  // Install skills that don't already exist in the project
-  const skillsDir = `${cwd}/.claude/skills`
+const setupSkills = async (skillsDir) => {
   const packageSkillsDir = `${__dirname}/../skills`
 
   if (await exists(packageSkillsDir)) {
@@ -81,10 +80,53 @@ if (hasClaudeCode) {
           console.log(`Skill '${skill}' already exists, skipping`)
         } else {
           await $`cp -r ${packageSkillsDir}/${skill} ${targetSkillDir}`
-          console.log(`Installed skill '${skill}' to .claude/skills/`)
+          console.log(`Installed skill '${skill}' to ${skillsDir}/`)
         }
       }),
     )
   }
-  // TODO: .codex/skills, .cursor/skills if needed
 }
+
+if (hasClaudeCode) {
+  // Install skills that don't already exist in the project
+  await setupSkills(`${cwd}/.claude/skills`)
+}
+
+if (hasAgentsMd) {
+  // Handle codex skills
+  // Currently no signal file for codex skills, so we just check if AGENTS.md exists
+  await setupSkills(`${cwd}/.codex/skills`)
+
+  const defaultCodexMcpConfig = {
+    mcp_servers: {
+      'bricks-project': projectMcpServer,
+    },
+  }
+
+  const handleCodexMcpConfigOverride = async (mcpConfigPath: string) => {
+    let mcpConfig: { mcp_servers: Record<string, typeof projectMcpServer> } | null = null
+    if (await exists(mcpConfigPath)) {
+      const configStr = await readFile(mcpConfigPath, 'utf-8')
+      try {
+        mcpConfig = TOML.parse(configStr)
+        if (!mcpConfig?.mcp_servers) throw new Error('mcp_servers is not defined')
+        mcpConfig.mcp_servers['bricks-project'] = projectMcpServer
+      } catch (e) {
+        mcpConfig = defaultCodexMcpConfig
+      }
+    } else {
+      mcpConfig = defaultCodexMcpConfig
+    }
+
+    await writeFile(mcpConfigPath, `${TOML.stringify(mcpConfig, null, 2)}\n`)
+
+    console.log(`Updated ${mcpConfigPath}`)
+  }
+
+  // Setup MCP config (.codex/config.toml)
+  const codexConfigPath = `${cwd}/.codex/config.toml`
+  await handleCodexMcpConfigOverride(codexConfigPath)
+}
+
+// TODO: .cursor/skills if needed
+// TODO: User setting in application.json to avoid unnecessary skills/config setup

package/types/generators/RealtimeTranscription.ts

@@ -41,10 +41,10 @@ Default property:
   "audioSliceSec": 30,
   "audioMinSec": 1,
   "maxSlicesInMemory": 5,
+  "transcribeProcessingPauseMs": 500,
+  "initTranscribeAfterMs": 500,
   "vadStrategy": "use-preset",
   "vadPreset": "default",
-  "autoSliceOnSpeechEnd": true,
-  "autoSliceThreshold": 2,
   "initialPrompt": "",
   "promptPreviousSlices": false,
   "saveAudio": true,
@@ -73,6 +73,10 @@ Default property:
     audioMinSec?: number | DataLink
     /* Maximum number of slices to keep in memory */
     maxSlicesInMemory?: number | DataLink
+    /* Transcribe processing interval in milliseconds */
+    transcribeProcessingPauseMs?: number | DataLink
+    /* Transcribe processing init after pause in milliseconds */
+    initTranscribeAfterMs?: number | DataLink
     /* VAD Strategy */
     vadStrategy?: 'use-preset' | 'use-generator-options' | DataLink
     /* VAD preset configuration */
@@ -86,10 +90,6 @@ Default property:
       | 'meeting'
       | 'noisy-environment'
       | DataLink
-    /* Auto slice on speech end */
-    autoSliceOnSpeechEnd?: boolean | DataLink
-    /* Auto slice threshold in seconds */
-    autoSliceThreshold?: number | DataLink
     /* Initial prompt for transcription */
     initialPrompt?: string | DataLink
     /* Include previous slices in prompt */
@@ -128,6 +128,8 @@ Default property:
     onStatusChange?: Array<EventAction>
     /* Event triggered when statistics update */
     onStatsUpdate?: Array<EventAction>
+    /* Event triggered when slice transcription is stabilized */
+    onStabilized?: Array<EventAction>
     /* Event triggered when transcription ends */
     onEnd?: Array<EventAction>
   }
@@ -140,12 +142,16 @@ Default property:
     results?: () => Data
     /* Current transcription result text */
     resultText?: () => Data
+    /* Last stabilized transcription result segment */
+    lastStabilizedSegment?: () => Data
     /* Current statistics */
     statistics?: () => Data
     /* Latest transcribe event */
     lastTranscribeEvent?: () => Data
     /* Latest VAD event */
     lastVadEvent?: () => Data
+    /* Stabilized transcription text from completed slices */
+    stabilizedText?: () => Data
     /* Audio output file path (auto-generated when saving audio) */
     audioOutputPath?: () => Data
   }
@@ -170,9 +176,11 @@ export type GeneratorRealtimeTranscription = Generator &
                     | 'isTranscribing'
                     | 'results'
                     | 'resultText'
+                    | 'lastStabilizedSegment'
                     | 'statistics'
                     | 'lastTranscribeEvent'
                     | 'lastVadEvent'
+                    | 'stabilizedText'
                     | 'audioOutputPath'
                   value: any
                 }

package/types/generators/SpeechToTextGgml.ts

@@ -80,76 +80,9 @@ export type GeneratorSpeechInferenceActionTranscribeData = ActionWithParams & {
   >
 }
 
-/* [Deprecated] Transcribe microphone audio source */
-export type GeneratorSpeechInferenceActionTranscribeRealtime = ActionWithParams & {
-  __actionName: 'GENERATOR_SPEECH_INFERENCE_TRANSCRIBE_REALTIME'
-  params?: Array<
-    | {
-        input: 'prompt'
-        value?: string | DataLink | EventProperty
-        mapping?: string
-      }
-    | {
-        input: 'beamSize'
-        value?: number | DataLink | EventProperty
-        mapping?: string
-      }
-    | {
-        input: 'language'
-        value?: string | DataLink | EventProperty
-        mapping?: string
-      }
-    | {
-        input: 'translate'
-        value?: boolean | DataLink | EventProperty
-        mapping?: string
-      }
-    | {
-        input: 'realtimeAudioSeconds'
-        value?: number | DataLink | EventProperty
-        mapping?: string
-      }
-    | {
-        input: 'realtimeAudioSliceSeconds'
-        value?: number | DataLink | EventProperty
-        mapping?: string
-      }
-    | {
-        input: 'realtimeAudioMinSeconds'
-        value?: number | DataLink | EventProperty
-        mapping?: string
-      }
-    | {
-        input: 'realtimeSaveAudio'
-        value?: boolean | DataLink | EventProperty
-        mapping?: string
-      }
-    | {
-        input: 'realtimeVadEnabled'
-        value?: boolean | DataLink | EventProperty
-        mapping?: string
-      }
-    | {
-        input: 'realtimeVadMs'
-        value?: number | DataLink | EventProperty
-        mapping?: string
-      }
-    | {
-        input: 'realtimeVadThold'
-        value?: number | DataLink | EventProperty
-        mapping?: string
-      }
-    | {
-        input: 'realtimeVadFreqThold'
-        value?: number | DataLink | EventProperty
-        mapping?: string
-      }
-  >
-}
-
-/* [Deprecated] Stop transcribing microphone audio source */
-export type GeneratorSpeechInferenceActionTranscribeRealtimeStop = Action & {
-  __actionName: 'GENERATOR_SPEECH_INFERENCE_TRANSCRIBE_REALTIME_STOP'
+/* Stop current transcription */
+export type GeneratorSpeechInferenceActionTranscribeStop = Action & {
+  __actionName: 'GENERATOR_SPEECH_INFERENCE_TRANSCRIBE_STOP'
 }
 
 /* Clear downloaded files (model, audio) & current jobs */
@@ -172,14 +105,7 @@ Default property:
   "modelUseCoreML": false,
   "modelUseGPU": true,
   "modelUseFlashAttn": false,
-  "inferLanguage": "Auto",
-  "inferRealtimeAudioSeconds": 30,
-  "inferRealtimeAudioSliceSeconds": 30,
-  "inferRealtimeSaveAudio": false,
-  "inferRealtimeVadEnabled": false,
-  "inferRealtimeVadMs": 2000,
-  "inferRealtimeVadThold": 0.6,
-  "inferRealtimeVadFreqThold": 100
+  "inferLanguage": "Auto"
 }
   */
   property?: {
@@ -381,23 +307,6 @@ Default property:
     inferFileUrl?: string | DataLink
     /* MD5 of file to be inferred */
     inferFileMd5?: string | DataLink
-    /* Record duration in seconds. For performance, the value less than 30 seconds is recommended. */
-    inferRealtimeAudioSeconds?: number | DataLink
-    /* Optimize audio transcription performance by slicing audio samples when `Realtime Audio Seconds` > 30. */
-    inferRealtimeAudioSliceSeconds?: number | DataLink
-    /* Min duration of audio to start transcribe. Min: 0.5 seconds, Max: `Realtime Audio Slice Seconds`, Default: 1 second */
-    inferRealtimeAudioMinSeconds?: number | DataLink
-    /* Save recorded audio to file, the file path will be output to outlet. */
-    inferRealtimeSaveAudio?: boolean | DataLink
-    /* Start transcribe on recording when the audio volume is greater than the threshold by using VAD (Voice Activity Detection).  
- The first VAD will be triggered after 2 second of recording. */
-    inferRealtimeVadEnabled?: boolean | DataLink
-    /* The length of the collected audio is used for VAD. (ms) */
-    inferRealtimeVadMs?: number | DataLink
-    /* VAD threshold */
-    inferRealtimeVadThold?: number | DataLink
-    /* Frequency to apply High-pass filter in VAD */
-    inferRealtimeVadFreqThold?: number | DataLink
     /* Buttress connection settings for remote inference */
     buttressConnectionSettings?:
       | DataLink

package/types/generators/VadOnnx.ts

@@ -0,0 +1,201 @@
+/* Auto generated by build script */
+import type { SwitchCondInnerStateCurrentCanvas, SwitchCondData, SwitchDef } from '../switch'
+import type { Data, DataLink } from '../data'
+import type {
+  Brick,
+  Generator,
+  EventAction,
+  ActionWithDataParams,
+  ActionWithParams,
+  Action,
+  EventProperty,
+} from '../common'
+
+/* Load the model */
+export type GeneratorVadInferenceOnnxActionLoadModel = Action & {
+  __actionName: 'GENERATOR_ONNX_VAD_LOAD_MODEL'
+}
+
+/* Detect speech in audio file */
+export type GeneratorVadInferenceOnnxActionDetectFile = ActionWithParams & {
+  __actionName: 'GENERATOR_ONNX_VAD_DETECT_FILE'
+  params?: Array<
+    | {
+        input: 'fileUrl'
+        value?: string | DataLink | EventProperty
+        mapping?: string
+      }
+    | {
+        input: 'threshold'
+        value?: number | DataLink | EventProperty
+        mapping?: string
+      }
+    | {
+        input: 'minSpeechDurationMs'
+        value?: number | DataLink | EventProperty
+        mapping?: string
+      }
+    | {
+        input: 'minSilenceDurationMs'
+        value?: number | DataLink | EventProperty
+        mapping?: string
+      }
+    | {
+        input: 'maxSpeechDurationS'
+        value?: number | DataLink | EventProperty
+        mapping?: string
+      }
+    | {
+        input: 'speechPadMs'
+        value?: number | DataLink | EventProperty
+        mapping?: string
+      }
+  >
+}
+
+/* Detect speech in audio data */
+export type GeneratorVadInferenceOnnxActionDetectData = ActionWithParams & {
+  __actionName: 'GENERATOR_ONNX_VAD_DETECT_DATA'
+  params?: Array<
+    | {
+        input: 'data'
+        value?: any | EventProperty
+        mapping?: string
+      }
+    | {
+        input: 'threshold'
+        value?: number | DataLink | EventProperty
+        mapping?: string
+      }
+    | {
+        input: 'minSpeechDurationMs'
+        value?: number | DataLink | EventProperty
+        mapping?: string
+      }
+    | {
+        input: 'minSilenceDurationMs'
+        value?: number | DataLink | EventProperty
+        mapping?: string
+      }
+    | {
+        input: 'maxSpeechDurationS'
+        value?: number | DataLink | EventProperty
+        mapping?: string
+      }
+    | {
+        input: 'speechPadMs'
+        value?: number | DataLink | EventProperty
+        mapping?: string
+      }
+  >
+}
+
+/* Clean cache */
+export type GeneratorVadInferenceOnnxActionCleanCache = Action & {
+  __actionName: 'GENERATOR_ONNX_VAD_CLEAN_CACHE'
+}
+
+/* Release context */
+export type GeneratorVadInferenceOnnxActionReleaseContext = Action & {
+  __actionName: 'GENERATOR_ONNX_VAD_RELEASE_CONTEXT'
+}
+
+interface GeneratorVadInferenceOnnxDef {
+  /*
+Default property:
+{
+  "modelType": "auto",
+  "detectThreshold": 0.5,
+  "detectMinSpeechDurationMs": 250,
+  "detectMinSilenceDurationMs": 100,
+  "detectMaxSpeechDurationS": 30,
+  "detectSpeechPadMs": 30,
+  "executionMode": "sequential"
+}
+  */
+  property?: {
+    /* Initialize the VAD context on generator initialization */
+    init?: boolean | DataLink
+    /* VAD model */
+    model?: string | DataLink
+    /* Model type (auto-detected from config.json) */
+    modelType?: string | DataLink
+    /* Quantize type */
+    quantizeType?:
+      | 'auto'
+      | 'none'
+      | 'fp16'
+      | 'q8'
+      | 'int8'
+      | 'uint8'
+      | 'q4'
+      | 'bnb4'
+      | 'q4f16'
+      | DataLink
+    /* Speech probability threshold (0.0-1.0) */
+    detectThreshold?: number | DataLink
+    /* Minimum speech duration in milliseconds */
+    detectMinSpeechDurationMs?: number | DataLink
+    /* Minimum silence duration in milliseconds */
+    detectMinSilenceDurationMs?: number | DataLink
+    /* Maximum speech duration in seconds */
+    detectMaxSpeechDurationS?: number | DataLink
+    /* Padding around speech segments in milliseconds */
+    detectSpeechPadMs?: number | DataLink
+    /* Executor candidates, descending order of priority  
+  Default will be xnnpack, wasm, cpu */
+    executors?:
+      | Array<'qnn' | 'dml' | 'nnapi' | 'xnnpack' | 'coreml' | 'cpu' | 'wasm' | 'webgpu' | DataLink>
+      | DataLink
+    /* Execution mode  
+  Usually when the model has many branches, setting this option to `parallel` will give you better performance. */
+    executionMode?: 'sequential' | 'parallel' | DataLink
+    /* QNN backend */
+    qnnBackend?: 'HTP' | 'HTA' | 'DSP' | 'GPU' | 'CPU' | DataLink
+    /* Enable FP16 for QNN HTP */
+    qnnHtpEnableFp16?: boolean | DataLink
+    /* Enable QNN debug */
+    qnnEnableDebug?: boolean | DataLink
+  }
+  events?: {
+    /* Event triggered when context state changes */
+    onContextStateChange?: Array<EventAction>
+    /* Event triggered when error occurs */
+    onError?: Array<EventAction>
+    /* Event triggered when got detection result */
+    onDetected?: Array<EventAction>
+  }
+  outlets?: {
+    /* Context state */
+    contextState?: () => Data
+    /* Is detecting */
+    isDetecting?: () => Data
+    /* Detection segments result */
+    detectionSegments?: () => Data
+    /* Detection details */
+    detectionDetails?: () => Data
+  }
+}
+
+/* Local Voice Activity Detection (VAD) inference based on [transformers.js](https://huggingface.co/docs/transformers.js)  
+ You can use any compatible VAD model from HuggingFace (Silero VAD, smart-turn, etc.) */
+export type GeneratorVadInferenceOnnx = Generator &
+  GeneratorVadInferenceOnnxDef & {
+    templateKey: 'GENERATOR_ONNX_VAD'
+    switches: Array<
+      SwitchDef &
+        GeneratorVadInferenceOnnxDef & {
+          conds?: Array<{
+            method: '==' | '!=' | '>' | '<' | '>=' | '<='
+            cond:
+              | SwitchCondInnerStateCurrentCanvas
+              | SwitchCondData
+              | {
+                  __typename: 'SwitchCondInnerStateOutlet'
+                  outlet: 'contextState' | 'isDetecting' | 'detectionSegments' | 'detectionDetails'
+                  value: any
+                }
+          }>
+        }
+    >
+  }

package/types/generators/VadTraditional.ts

@@ -0,0 +1,123 @@
+/* Auto generated by build script */
+import type { SwitchCondInnerStateCurrentCanvas, SwitchCondData, SwitchDef } from '../switch'
+import type { Data, DataLink } from '../data'
+import type {
+  Brick,
+  Generator,
+  EventAction,
+  ActionWithDataParams,
+  ActionWithParams,
+  Action,
+  EventProperty,
+} from '../common'
+
+/* Detect speech in audio file */
+export type GeneratorVadInferenceTraditionalActionDetectFile = ActionWithParams & {
+  __actionName: 'GENERATOR_TRADITIONAL_VAD_DETECT_FILE'
+  params?: Array<{
+    input: 'fileUrl'
+    value?: string | DataLink | EventProperty
+    mapping?: string
+  }>
+}
+
+/* Detect speech in audio data stream */
+export type GeneratorVadInferenceTraditionalActionDetectData = ActionWithParams & {
+  __actionName: 'GENERATOR_TRADITIONAL_VAD_DETECT_DATA'
+  params?: Array<{
+    input: 'data'
+    value?: any | EventProperty
+    mapping?: string
+  }>
+}
+
+interface GeneratorVadInferenceTraditionalDef {
+  /*
+Default property:
+{
+  "detectVocalFreqMin": 75,
+  "detectVocalFreqMax": 900,
+  "detectThreshold": 0.5,
+  "detectMinSpeechDurationMs": 250,
+  "detectMinSilenceDurationMs": 100,
+  "detectMaxSpeechDurationS": 30,
+  "detectSpeechPadMs": 30
+}
+  */
+  property?: {
+    /* Minimum vocal frequency in Hz */
+    detectVocalFreqMin?: number | DataLink
+    /* Maximum vocal frequency in Hz */
+    detectVocalFreqMax?: number | DataLink
+    /* Volume threshold in dB */
+    detectVolumeThreshold?: number | DataLink
+    /* Speech probability threshold (0.0-1.0) - maps to frequency clarity */
+    detectThreshold?: number | DataLink
+    /* Minimum speech duration in milliseconds */
+    detectMinSpeechDurationMs?: number | DataLink
+    /* Minimum silence duration in milliseconds */
+    detectMinSilenceDurationMs?: number | DataLink
+    /* Maximum speech duration in seconds */
+    detectMaxSpeechDurationS?: number | DataLink
+    /* Padding around speech segments in milliseconds */
+    detectSpeechPadMs?: number | DataLink
+    /* The file URL or path to be analyzed */
+    detectFileUrl?: string | DataLink
+    /* MD5 of file to be analyzed */
+    detectFileMd5?: string | DataLink
+  }
+  events?: {
+    /* Event triggered when context state changes */
+    onContextStateChange?: Array<EventAction>
+    /* Event triggered when detection result is available */
+    onDetected?: Array<EventAction>
+    /* Event triggered when error occurs */
+    onError?: Array<EventAction>
+  }
+  outlets?: {
+    /* Context state */
+    contextState?: () => Data
+    /* Is detecting */
+    isDetecting?: () => Data
+    /* Is speaking (real-time) */
+    isSpeaking?: () => Data
+    /* Detection segments result */
+    detectionSegments?: () => Data
+    /* Current volume in dB */
+    currentVolume?: () => Data
+    /* Current frequency clarity (0-1) */
+    currentClarity?: () => Data
+    /* Current detected frequency in Hz */
+    currentFrequency?: () => Data
+  }
+}
+
+/* Traditional Voice Activity Detection (VAD) using pitch detection and RMS volume analysis  
+ No model download required - pure algorithmic approach */
+export type GeneratorVadInferenceTraditional = Generator &
+  GeneratorVadInferenceTraditionalDef & {
+    templateKey: 'GENERATOR_TRADITIONAL_VAD'
+    switches: Array<
+      SwitchDef &
+        GeneratorVadInferenceTraditionalDef & {
+          conds?: Array<{
+            method: '==' | '!=' | '>' | '<' | '>=' | '<='
+            cond:
+              | SwitchCondInnerStateCurrentCanvas
+              | SwitchCondData
+              | {
+                  __typename: 'SwitchCondInnerStateOutlet'
+                  outlet:
+                    | 'contextState'
+                    | 'isDetecting'
+                    | 'isSpeaking'
+                    | 'detectionSegments'
+                    | 'currentVolume'
+                    | 'currentClarity'
+                    | 'currentFrequency'
+                  value: any
+                }
+          }>
+        }
+    >
+  }

package/types/generators/index.ts

@@ -33,9 +33,11 @@ export * from './ThermalPrinter'
 export * from './SqLite'
 export * from './McpServer'
 export * from './Mcp'
+export * from './VadTraditional'
 export * from './TextToSpeechOnnx'
 export * from './LlmOnnx'
 export * from './SpeechToTextOnnx'
+export * from './VadOnnx'
 export * from './SpeechToTextGgml'
 export * from './VadGgml'
 export * from './RealtimeTranscription'

package/utils/event-props.ts

@@ -733,6 +733,18 @@ export const templateEventPropsMap = {
       'GENERATOR_MCP_ERROR_MESSAGE', // type: string
     ],
   },
+  GENERATOR_TRADITIONAL_VAD: {
+    onContextStateChange: [
+      'GENERATOR_TRADITIONAL_VAD_CONTEXT_STATE', // type: string
+    ],
+    onDetected: [
+      'GENERATOR_TRADITIONAL_VAD_DETECTION_SEGMENTS', // type: array
+      'GENERATOR_TRADITIONAL_VAD_DETECTION_TIME', // type: number
+    ],
+    onError: [
+      'GENERATOR_TRADITIONAL_VAD_ERROR', // type: string
+    ],
+  },
   GENERATOR_TTS: {
     onContextStateChange: [
       'GENERATOR_TTS_CONTEXT_STATE', // type: string
@@ -767,6 +779,18 @@ export const templateEventPropsMap = {
       'GENERATOR_ONNX_STT_ERROR', // type: string
     ],
   },
+  GENERATOR_ONNX_VAD: {
+    onContextStateChange: [
+      'GENERATOR_ONNX_VAD_CONTEXT_STATE', // type: string
+    ],
+    onError: [
+      'GENERATOR_ONNX_VAD_ERROR', // type: string
+    ],
+    onDetected: [
+      'GENERATOR_ONNX_VAD_DETECTION_SEGMENTS', // type: array
+      'GENERATOR_ONNX_VAD_DETECTION_TIME', // type: number
+    ],
+  },
   GENERATOR_SPEECH_INFERENCE: {
     onContextStateChange: [
       'GENERATOR_SPEECH_INFERENCE_CONTEXT_STATE', // type: string
@@ -823,6 +847,9 @@ export const templateEventPropsMap = {
       'GENERATOR_REALTIME_TRANSCRIPTION_STATS_TIMESTAMP', // type: number
       'GENERATOR_REALTIME_TRANSCRIPTION_STATS', // type: object
     ],
+    onStabilized: [
+      'GENERATOR_REALTIME_TRANSCRIPTION_STABILIZED_TEXT', // type: string
+    ],
     onEnd: [
       'GENERATOR_REALTIME_TRANSCRIPTION_END_RESULTS', // type: array
       'GENERATOR_REALTIME_TRANSCRIPTION_END_AUDIO_OUTPUT_PATH', // type: string