@fugood/bricks-project 2.23.0 → 2.23.2

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',

package/compile/index.ts

@@ -1,5 +1,6 @@
 /* eslint-disable no-underscore-dangle -- Uses __typename, __actionName, etc. for type system */
-import _ from 'lodash'
+import snakeCase from 'lodash/snakeCase'
+import omit from 'lodash/omit'
 import { parse as parseAST } from 'acorn'
 import type { ExportNamedDeclaration, FunctionDeclaration } from 'acorn'
 import escodegen from 'escodegen'
@@ -70,7 +71,7 @@ const compileEventActionValue = (templateKey, eventKey, value, errorReference) =
 }
 
 const convertOutletKey = (templateKey: string, key: string) =>
-  `${templateKey}_${_.snakeCase(key).toUpperCase()}`
+  `${templateKey}_${snakeCase(key).toUpperCase()}`
 
 const compileOutlets = (
   templateKey: string,
@@ -84,7 +85,7 @@ const compileOutlets = (
   }, {})
 
 const convertEventKey = (templateKey: string, key: string) =>
-  `${templateKey ? `${templateKey}_` : ''}${_.snakeCase(key).toUpperCase()}`
+  `${templateKey ? `${templateKey}_` : ''}${snakeCase(key).toUpperCase()}`
 
 const basicAnimationEvents = ['show', 'standby', 'breatheStart']
 
@@ -421,8 +422,9 @@ 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,
@@ -459,7 +461,7 @@ export const compile = async (app: Application) => {
               property: animationDef.property,
               type: animationTypeMap[animationDef.config.__type],
               config: compileProperty(
-                _.omit(animationDef.config, '__type'),
+                omit(animationDef.config, '__type'),
                 `(animation: ${animation.id}, subspace ${subspace.id})`,
               ),
             }
@@ -781,6 +783,7 @@ export const compile = async (app: Application) => {
             title: dataCalc.title,
             description: dataCalc.description,
           }
+          if (dataCalc.triggerMode) calc.trigger_type = dataCalc.triggerMode
           if (dataCalc.__typename === 'DataCalculationMap') {
             calc.type = 'general'
             const mapCalc = dataCalc as DataCalculationMap
@@ -915,6 +918,7 @@ export const compile = async (app: Application) => {
               note: scriptCalc.note,
               code,
               enable_async: scriptCalc.enableAsync,
+              trigger_mode: scriptCalc.triggerMode,
               inputs: scriptCalc.inputs.reduce((acc, input) => {
                 acc[input.data().id] = input.key
                 return acc
@@ -967,6 +971,7 @@ 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

@@ -1,6 +1,6 @@
 {
   "name": "@fugood/bricks-project",
-  "version": "2.23.0",
+  "version": "2.23.2",
   "main": "index.ts",
   "scripts": {
     "build": "bun scripts/build.js"
@@ -8,6 +8,7 @@
   "dependencies": {
     "@fugood/bricks-cli": "^2.23.0",
     "@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",
@@ -17,6 +18,5 @@
     "fuse.js": "^7.0.0",
     "lodash": "^4.17.4",
     "uuid": "^8.3.1"
-  },
-  "gitHead": "398352b9923f97e914ac60acab519ca014aa6fb8"
+  }
 }

package/skills/bricks-project/rules/automations.md

@@ -4,15 +4,16 @@ E2E testing and scheduled execution for BRICKS applications. Simulates user beha
 
 ## Automation Types
 
-| Type | Description |
-|------|-------------|
-| `launch` | Run on application launch (restarts app when run manually) |
-| `anytime` | Execute anytime via manual trigger |
-| `cron` | Scheduled execution using crontab expressions |
+| Type      | Description                                                |
+| --------- | ---------------------------------------------------------- |
+| `launch`  | Run on application launch (restarts app when run manually) |
+| `anytime` | Execute anytime via manual trigger                         |
+| `cron`    | Scheduled execution using crontab expressions              |
 
 ## Simulation Actions
 
 Automations can simulate:
+
 - **Brick Press**: Tap/click on bricks
 - **Key Events**: Key up/down for keyboard input
 - **HTTP Request**: API calls
@@ -21,6 +22,7 @@ Automations can simulate:
 ## Assertions
 
 Automations can validate:
+
 - **Brick Exists**: Check if brick is rendered
 - **Event Triggered**: Verify event from Brick/Generator/Canvas
 - **Canvas Changed**: Confirm canvas navigation
@@ -100,25 +102,58 @@ const testLoginFlow: AutomationTest = {
 
 ## Test Methods
 
-| Method | Signature | Description |
-|--------|-----------|-------------|
-| `brick_press` | `[subspace, brick, options?]` | Simulate brick press |
-| `brick_exists` | `[subspace, brick, frame?]` | Check brick exists |
-| `wait_until_brick_exists` | `[subspace, brick, timeout?, frame?]` | Wait for brick |
-| `wait_until_event_trigger` | `[subspace, sender, eventKey, timeout?]` | Wait for event |
-| `wait_until_canvas_change` | `[subspace, canvas, timeout?]` | Wait for canvas |
-| `keydown` | `[keyCode, pressedKey?, flags?]` | Key down event |
-| `keyup` | `[keyCode, pressedKey?, flags?]` | Key up event |
-| `http_request` | `[url, options?]` | HTTP request |
-| `assert_property` | `[subspace, property, value]` | Assert data value |
-| `wait_until_property_change` | `[subspace, property, value, timeout?]` | Wait for value |
-| `execute_action` | `[subspace, handler, action, params?, options?]` | Execute action |
-| `match_screenshot` | `[name, threshold?, maxRetry?]` | Screenshot compare |
-| `delay` | `[subspace?, property?, defaultValue?]` | Delay execution |
+| Method                       | Signature                                        | Description          |
+| ---------------------------- | ------------------------------------------------ | -------------------- |
+| `brick_press`                | `[subspace, brick, options?]`                    | Simulate brick press |
+| `brick_exists`               | `[subspace, brick, frame?]`                      | Check brick exists   |
+| `wait_until_brick_exists`    | `[subspace, brick, timeout?, frame?]`            | Wait for brick       |
+| `wait_until_event_trigger`   | `[subspace, sender, eventKey, timeout?]`         | Wait for event       |
+| `wait_until_canvas_change`   | `[subspace, canvas, timeout?]`                   | Wait for canvas      |
+| `keydown`                    | `[keyCode, pressedKey?, flags?]`                 | Key down event       |
+| `keyup`                      | `[keyCode, pressedKey?, flags?]`                 | Key up event         |
+| `http_request`               | `[url, options?]`                                | HTTP request         |
+| `assert_property`            | `[subspace, property, value]`                    | Assert data value    |
+| `wait_until_property_change` | `[subspace, property, value, timeout?]`          | Wait for value       |
+| `execute_action`             | `[subspace, handler, action, params?, options?]` | Execute action       |
+| `match_screenshot`           | `[name, threshold?, maxRetry?]`                  | Screenshot compare   |
+| `delay`                      | `[subspace?, property?, defaultValue?]`          | Delay execution      |
+
+### execute_action Params
+
+The `params` object in `execute_action` uses **runtime event property keys** from `event-props.ts`, NOT the action config `input` names from type definitions.
+
+```typescript
+// CORRECT — use runtime event property key
+run: ['execute_action', () => subspace0, bricks.bInput.id, 'BRICK_TEXT_INPUT_SET_TEXT',
+  { BRICK_TEXT_INPUT_TEXT: 'hello' }]
+
+// WRONG — action config input name doesn't work in automation
+run: ['execute_action', () => subspace0, bricks.bInput.id, 'BRICK_TEXT_INPUT_SET_TEXT',
+  { text: 'hello' }]
+```
+
+Reference `event-props.ts` for the correct runtime keys (e.g., `BRICK_TEXT_INPUT_TEXT`, `GENERATOR_MQTT_PAYLOAD`).
+
+### Prefer UI Interactions Over Direct Generator Calls
+
+For realistic E2E testing, prefer simulating user actions (set text input + press button) over calling generator actions directly:
+
+```typescript
+// GOOD — simulates real user behavior
+{ run: ['execute_action', () => sub, bricks.bInput.id, 'BRICK_TEXT_INPUT_SET_TEXT',
+    { BRICK_TEXT_INPUT_TEXT: 'hello' }] },
+{ run: ['brick_press', () => sub, () => bricks.bSendBtn] },
+{ run: ['wait_until_property_change', () => sub, () => data.dPayload, 'hello', 10000] },
+
+// AVOID — bypasses UI, doesn't test the full flow
+{ run: ['execute_action', () => sub, generators.gClient.id, 'GENERATOR_MQTT_PUBLISH',
+    { topic: 'test', payload: 'hello', qos: '0' }] },
+```
 
 ## Recording Automations
 
 In BRICKS Editor Preview mode:
+
 1. Perform operations normally
 2. Open menu (right-bottom corner)
 3. Select "Record Events as Automation"
@@ -127,12 +162,15 @@ In BRICKS Editor Preview mode:
 ## Running Automations
 
 ### Manual Run
+
 `Menu` → `Automations` → Select automation → `Run`
 
 ### On Launch
+
 `Bind Device` → `Select Automation` (only `launch` or `cron` types)
 
 ### Scheduled (Cron)
+
 `Bind Device` → `Cron Automation` (allows multi-select)
 
 Use [crontab.guru](https://crontab.guru) to build cron expressions.
@@ -156,6 +194,7 @@ Visual regression testing with screenshot comparison:
 ```
 
 Screenshots can be stored:
+
 - Local file system
 - Media Flow workspace
 
@@ -165,11 +204,18 @@ First run captures baseline. Use "Run with Update" to update baseline.
 
 Automations work with Modules. Use Manual Run in Preview mode for module testing.
 
+## Important Notes
+
+- **Automation map key**: Always use `'AUTOMATION_MAP_DEFAULT'` as the automation map ID (not `makeId()`). The preview test runner reads from `automationMap['AUTOMATION_MAP_DEFAULT']?.map`.
+- **Valid makeId types**: Use `'test'` for AutomationTest, `'test_case'` for TestCase, `'test_var'` for TestVariable. Do NOT use `'automation_test'` or `'automation_test_map'`.
+- **handler in execute_action**: Pass the entity's `.id` string (e.g., `bricks.bInput.id`), not a getter function.
+
 ## Best Practices
 
 1. **Test culture**: Create automations for every significant flow
 2. **CI/CD integration**: Use `launch` automations for deployment validation
-3. **Incremental waits**: Use `EXPECT_*` steps with appropriate timeouts
+3. **Incremental waits**: Use `wait_until_property_change` with appropriate timeouts
 4. **Visual testing**: Add screenshot comparisons for critical UI states
 5. **Cron monitoring**: Schedule health checks for production displays
 6. **Isolation**: Each automation should be independent and idempotent
+7. **UI-first testing**: Simulate real user interactions (text input, button press) rather than calling generators directly

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/tools/preview-main.mjs

@@ -42,13 +42,14 @@ let config = JSON.parse(await readFile(`${cwd}/.bricks/build/application-config.
 let testId = values['test-id'] || null
 if (!testId && values['test-title-like']) {
   const titleLike = values['test-title-like'].toLowerCase()
-  const testMap = config.test_map || {}
-  const found = Object.entries(testMap).find(([, test]) =>
-    test.title?.toLowerCase().includes(titleLike),
-  )
-  if (found) {
-    ;[testId] = found
-  } else {
+  const automationMap = config.automation_map || {}
+  const matchedEntry = Object.values(automationMap)
+    .flatMap((group) => Object.entries(group.map || {}))
+    .find(([, test]) => test.title?.toLowerCase().includes(titleLike))
+  if (matchedEntry) {
+    ;[testId] = matchedEntry
+  }
+  if (!testId) {
     throw new Error(`No automation found matching title: ${values['test-title-like']}`)
   }
 }

package/tools/preview.ts

@@ -2,7 +2,7 @@ import { $ } from 'bun'
 import { watch } from 'fs'
 import type { FSWatcher } from 'fs'
 import { parseArgs } from 'util'
-import { debounce } from 'lodash'
+import debounce from 'lodash/debounce'
 
 const { values } = parseArgs({
   args: Bun.argv,

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

@@ -147,6 +147,11 @@ export type GeneratorSpeechInferenceActionTranscribeRealtime = ActionWithParams
   >
 }
 
+/* Stop current transcription */
+export type GeneratorSpeechInferenceActionTranscribeStop = Action & {
+  __actionName: 'GENERATOR_SPEECH_INFERENCE_TRANSCRIBE_STOP'
+}
+
 /* [Deprecated] Stop transcribing microphone audio source */
 export type GeneratorSpeechInferenceActionTranscribeRealtimeStop = Action & {
   __actionName: 'GENERATOR_SPEECH_INFERENCE_TRANSCRIBE_REALTIME_STOP'

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/calc.ts

@@ -33,10 +33,11 @@ export const generateDataCalculationMapEditorInfo = (
   nodes.forEach((node) => {
     // Count and track inputs
     if ('inputs' in node) {
-      const inputs = node.inputs
-        .filter((input) => input !== null)
-        .map((input) => (Array.isArray(input) ? input.length : 1))
-        .reduce((sum, count) => sum + count, 0)
+      const inputs = node.inputs.reduce((count, input) => {
+        if (input === null) return count
+        if (Array.isArray(input)) return count + input.length
+        return count + 1
+      }, 0)
       inputCounts.set(node, inputs)
 
       // Track connections
@@ -59,10 +60,11 @@ export const generateDataCalculationMapEditorInfo = (
 
     // Count outputs
     if ('outputs' in node) {
-      const outputs = node.outputs
-        .filter((output) => output !== null)
-        .map((output) => (Array.isArray(output) ? output.length : 1))
-        .reduce((sum, count) => sum + count, 0)
+      const outputs = node.outputs.reduce((count, output) => {
+        if (output === null) return count
+        if (Array.isArray(output)) return count + output.length
+        return count + 1
+      }, 0)
       outputCounts.set(node, outputs)
     } else {
       outputCounts.set(node, 0)

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