elsabro 2.2.0 → 3.7.0

package/templates/voice-commands-config.json

@@ -0,0 +1,658 @@
+{
+  "$schema": "https://elsabro.dev/schemas/voice-config-v3.7.json",
+  "version": "3.7.0",
+  "name": "ELSABRO Voice Commands Configuration",
+  "description": "Complete configuration for voice commands and dictation system",
+
+  "voiceEngine": {
+    "enabled": true,
+    "autoStart": false,
+    "sampleRate": 16000,
+    "channels": 1,
+    "bufferSize": 4096,
+    "maxRecordingDuration": 30000,
+    "silenceTimeout": 1500,
+    "audioFormat": "pcm_f32le"
+  },
+
+  "asrProvider": {
+    "provider": "whisper-api",
+    "providers": {
+      "whisper-api": {
+        "enabled": true,
+        "model": "whisper-1",
+        "endpoint": "https://api.openai.com/v1/audio/transcriptions",
+        "apiKeyEnvVar": "OPENAI_API_KEY",
+        "maxFileSizeMB": 25,
+        "supportedFormats": ["flac", "m4a", "mp3", "mp4", "mpeg", "mpga", "oga", "ogg", "wav", "webm"],
+        "timeout": 30000
+      },
+      "whisper-local": {
+        "enabled": false,
+        "modelPath": "./models/whisper-large-v3-turbo.onnx",
+        "modelSize": "large-v3-turbo",
+        "device": "auto",
+        "computeType": "float16",
+        "beamSize": 5,
+        "vadFilter": true,
+        "wordTimestamps": true
+      },
+      "azure": {
+        "enabled": false,
+        "endpoint": "https://{region}.stt.speech.microsoft.com/speech/recognition/conversation/cognitiveservices/v1",
+        "apiKeyEnvVar": "AZURE_SPEECH_KEY",
+        "regionEnvVar": "AZURE_REGION",
+        "outputFormat": "detailed",
+        "profanityFilter": "raw"
+      },
+      "google": {
+        "enabled": false,
+        "endpoint": "https://speech.googleapis.com/v1/speech:recognize",
+        "apiKeyEnvVar": "GOOGLE_API_KEY",
+        "model": "latest_long",
+        "useEnhanced": true,
+        "enableWordTimeOffsets": true
+      }
+    },
+    "fallbackOrder": ["whisper-api", "whisper-local", "azure", "google"]
+  },
+
+  "language": {
+    "default": "es",
+    "autoDetect": true,
+    "supported": ["es", "en", "pt"],
+    "languageConfigs": {
+      "es": {
+        "name": "Spanish",
+        "nativeName": "Espanol",
+        "locale": "es-ES",
+        "whisperCode": "es",
+        "ttsVoice": "nova",
+        "azureVoice": "es-ES-ElviraNeural",
+        "googleVoice": "es-ES-Neural2-A",
+        "dateFormat": "DD/MM/YYYY",
+        "timeFormat": "HH:mm",
+        "numberFormat": "es-ES"
+      },
+      "en": {
+        "name": "English",
+        "nativeName": "English",
+        "locale": "en-US",
+        "whisperCode": "en",
+        "ttsVoice": "alloy",
+        "azureVoice": "en-US-JennyNeural",
+        "googleVoice": "en-US-Neural2-J",
+        "dateFormat": "MM/DD/YYYY",
+        "timeFormat": "h:mm A",
+        "numberFormat": "en-US"
+      },
+      "pt": {
+        "name": "Portuguese",
+        "nativeName": "Portugues",
+        "locale": "pt-BR",
+        "whisperCode": "pt",
+        "ttsVoice": "shimmer",
+        "azureVoice": "pt-BR-FranciscaNeural",
+        "googleVoice": "pt-BR-Neural2-A",
+        "dateFormat": "DD/MM/YYYY",
+        "timeFormat": "HH:mm",
+        "numberFormat": "pt-BR"
+      }
+    }
+  },
+
+  "wakeWord": {
+    "enabled": true,
+    "phrase": "hey elsabro",
+    "alternatives": ["ok elsabro", "elsabro"],
+    "sensitivity": 0.7,
+    "engine": "porcupine",
+    "engines": {
+      "porcupine": {
+        "accessKeyEnvVar": "PICOVOICE_ACCESS_KEY",
+        "keywordPath": "./models/hey_elsabro.ppn",
+        "modelPath": "./models/porcupine_params.pv",
+        "endpointDuration": 1.0
+      },
+      "custom": {
+        "modelPath": "./models/wake_word_custom.onnx",
+        "threshold": 0.8,
+        "windowSize": 1.5
+      }
+    },
+    "requireConsecutive": 2,
+    "cooldownMs": 2000,
+    "feedbackSound": true
+  },
+
+  "vad": {
+    "enabled": true,
+    "engine": "webrtc",
+    "engines": {
+      "webrtc": {
+        "mode": 3,
+        "frameDurationMs": 30,
+        "sampleRate": 16000
+      },
+      "silero": {
+        "modelPath": "./models/silero_vad.onnx",
+        "threshold": 0.5,
+        "minSpeechDurationMs": 250,
+        "minSilenceDurationMs": 100,
+        "windowSizeSamples": 1536,
+        "speechPadMs": 30
+      },
+      "energy": {
+        "threshold": 0.01,
+        "zeroCrossingThreshold": 0.1
+      }
+    },
+    "speechPaddingMs": 300,
+    "maxSpeechDurationMs": 30000
+  },
+
+  "noiseReduction": {
+    "enabled": true,
+    "noiseGateDb": -40,
+    "echoCancellation": true,
+    "autoGainControl": true,
+    "noiseSuppression": true,
+    "spectralSubtraction": {
+      "enabled": true,
+      "overSubtractionFactor": 2.0,
+      "spectralFloor": 0.01,
+      "smoothingFactor": 0.98
+    },
+    "calibration": {
+      "durationMs": 3000,
+      "autoCalibrate": true,
+      "recalibrateOnNoiseChange": true
+    }
+  },
+
+  "intentClassifier": {
+    "modelPath": "./models/elsabro-intent-v1.onnx",
+    "tokenizerPath": "./models/intent-tokenizer.json",
+    "confidenceThreshold": 0.7,
+    "fallbackToLLM": true,
+    "llmModel": "claude-3-5-sonnet-20241022",
+    "cacheResults": true,
+    "cacheTTLMs": 3600000,
+    "intents": {
+      "execute_plan": {
+        "patterns": [
+          "ejecuta el plan",
+          "corre el proyecto",
+          "inicia la ejecucion",
+          "run the plan",
+          "execute plan",
+          "start execution",
+          "executa o plano"
+        ],
+        "action": "/elsabro:execute",
+        "requiresConfirmation": false
+      },
+      "stop_execution": {
+        "patterns": [
+          "para",
+          "detente",
+          "stop",
+          "cancela",
+          "termina"
+        ],
+        "action": "/elsabro:stop",
+        "requiresConfirmation": false
+      },
+      "pause_execution": {
+        "patterns": [
+          "pausa",
+          "pause",
+          "espera",
+          "wait"
+        ],
+        "action": "/elsabro:pause",
+        "requiresConfirmation": false
+      },
+      "resume_execution": {
+        "patterns": [
+          "continua",
+          "resume",
+          "sigue",
+          "continue"
+        ],
+        "action": "/elsabro:resume",
+        "requiresConfirmation": false
+      },
+      "show_progress": {
+        "patterns": [
+          "muestra el progreso",
+          "como va",
+          "show progress",
+          "status"
+        ],
+        "action": "/elsabro:progress",
+        "requiresConfirmation": false
+      },
+      "create_task": {
+        "patterns": [
+          "crea una tarea",
+          "nueva tarea",
+          "create task",
+          "new task",
+          "add task"
+        ],
+        "action": "TaskCreate",
+        "requiresConfirmation": true,
+        "entities": ["task_description", "priority", "assignee"]
+      },
+      "list_tasks": {
+        "patterns": [
+          "lista las tareas",
+          "muestra tareas",
+          "list tasks",
+          "show tasks"
+        ],
+        "action": "TaskList",
+        "requiresConfirmation": false
+      },
+      "complete_task": {
+        "patterns": [
+          "completa la tarea",
+          "termina tarea",
+          "complete task",
+          "finish task"
+        ],
+        "action": "TaskComplete",
+        "requiresConfirmation": true,
+        "entities": ["task_id", "task_description"]
+      },
+      "open_file": {
+        "patterns": [
+          "abre el archivo",
+          "abrir",
+          "open file",
+          "open"
+        ],
+        "action": "FileOpen",
+        "requiresConfirmation": false,
+        "entities": ["file_path"]
+      },
+      "goto_line": {
+        "patterns": [
+          "ir a la linea",
+          "ve a linea",
+          "go to line",
+          "line"
+        ],
+        "action": "GoToLine",
+        "requiresConfirmation": false,
+        "entities": ["line_number"]
+      },
+      "search_code": {
+        "patterns": [
+          "busca",
+          "encuentra",
+          "search",
+          "find"
+        ],
+        "action": "Search",
+        "requiresConfirmation": false,
+        "entities": ["search_query"]
+      },
+      "switch_agent": {
+        "patterns": [
+          "cambia al agente",
+          "usa agente",
+          "switch to agent",
+          "use agent"
+        ],
+        "action": "AgentSwitch",
+        "requiresConfirmation": false,
+        "entities": ["agent_name"]
+      },
+      "list_agents": {
+        "patterns": [
+          "lista los agentes",
+          "que agentes hay",
+          "list agents",
+          "show agents"
+        ],
+        "action": "AgentList",
+        "requiresConfirmation": false
+      },
+      "show_logs": {
+        "patterns": [
+          "muestra los logs",
+          "ver logs",
+          "show logs",
+          "display logs"
+        ],
+        "action": "ShowLogs",
+        "requiresConfirmation": false
+      },
+      "show_errors": {
+        "patterns": [
+          "muestra los errores",
+          "hay errores",
+          "show errors",
+          "display errors"
+        ],
+        "action": "ShowErrors",
+        "requiresConfirmation": false
+      },
+      "help": {
+        "patterns": [
+          "ayuda",
+          "que puedo hacer",
+          "help",
+          "what can I do",
+          "commands"
+        ],
+        "action": "ShowHelp",
+        "requiresConfirmation": false
+      },
+      "calibrate": {
+        "patterns": [
+          "calibra",
+          "calibrar microfono",
+          "calibrate",
+          "calibrate mic"
+        ],
+        "action": "Calibrate",
+        "requiresConfirmation": false
+      }
+    },
+    "entities": {
+      "file_path": {
+        "type": "regex",
+        "patterns": [
+          "(?:archivo|file)\\s+([\\w\\-\\.\\/]+)",
+          "([\\w\\-]+\\/[\\w\\-\\.]+)"
+        ]
+      },
+      "line_number": {
+        "type": "regex",
+        "patterns": [
+          "(?:linea|line)\\s*(\\d+)",
+          "^(\\d+)$"
+        ]
+      },
+      "task_description": {
+        "type": "slot",
+        "extractAfter": ["para", "for", "de", "to"],
+        "maxLength": 200
+      },
+      "agent_name": {
+        "type": "enum",
+        "values": ["frontend", "backend", "fullstack", "devops", "qa", "architect", "pm"],
+        "fuzzyMatch": true
+      },
+      "priority": {
+        "type": "enum",
+        "values": ["alta", "high", "media", "medium", "baja", "low", "urgente", "urgent"],
+        "defaultValue": "medium"
+      },
+      "search_query": {
+        "type": "slot",
+        "extractAfter": ["busca", "search", "find", "encuentra"],
+        "maxLength": 100
+      }
+    }
+  },
+
+  "dictation": {
+    "enabled": true,
+    "streamingMode": true,
+    "interimResults": true,
+    "punctuationRestoration": {
+      "enabled": true,
+      "modelPath": "./models/punctuation-bert-multilingual.onnx",
+      "capitalize": true,
+      "addPeriods": true,
+      "addCommas": true,
+      "addQuestionMarks": true
+    },
+    "errorCorrection": {
+      "enabled": true,
+      "level": "basic",
+      "customDictionary": "./config/technical-terms.json",
+      "phoneticMatching": true
+    },
+    "codeDetection": {
+      "enabled": true,
+      "supportedLanguages": ["typescript", "javascript", "python", "sql", "bash", "go", "rust"],
+      "formatAsCodeBlock": true,
+      "highlightSyntax": true
+    },
+    "formatting": {
+      "newParagraphKeywords": ["nuevo parrafo", "new paragraph", "siguiente parrafo"],
+      "newLineKeywords": ["nueva linea", "new line", "enter"],
+      "deleteKeywords": ["borra eso", "delete that", "elimina"]
+    }
+  },
+
+  "tts": {
+    "enabled": true,
+    "provider": "openai",
+    "providers": {
+      "openai": {
+        "endpoint": "https://api.openai.com/v1/audio/speech",
+        "apiKeyEnvVar": "OPENAI_API_KEY",
+        "model": "tts-1",
+        "voices": ["alloy", "echo", "fable", "onyx", "nova", "shimmer"],
+        "defaultVoice": "nova",
+        "maxLength": 4096
+      },
+      "azure": {
+        "endpoint": "https://{region}.tts.speech.microsoft.com/cognitiveservices/v1",
+        "apiKeyEnvVar": "AZURE_SPEECH_KEY",
+        "regionEnvVar": "AZURE_REGION",
+        "outputFormat": "audio-16khz-128kbitrate-mono-mp3",
+        "voices": {
+          "es": "es-ES-ElviraNeural",
+          "en": "en-US-JennyNeural",
+          "pt": "pt-BR-FranciscaNeural"
+        }
+      },
+      "google": {
+        "endpoint": "https://texttospeech.googleapis.com/v1/text:synthesize",
+        "apiKeyEnvVar": "GOOGLE_API_KEY",
+        "audioEncoding": "MP3",
+        "voices": {
+          "es": "es-ES-Neural2-A",
+          "en": "en-US-Neural2-J",
+          "pt": "pt-BR-Neural2-A"
+        }
+      },
+      "system": {
+        "enabled": true,
+        "fallback": true
+      }
+    },
+    "settings": {
+      "speed": 1.0,
+      "pitch": 1.0,
+      "volume": 0.8,
+      "pauseBetweenSentencesMs": 300
+    },
+    "caching": {
+      "enabled": true,
+      "maxCacheSizeMB": 100,
+      "ttlMs": 86400000,
+      "cacheDir": "./.cache/tts"
+    }
+  },
+
+  "audioFeedback": {
+    "enabled": true,
+    "tones": {
+      "activation": {
+        "frequency": 800,
+        "duration": 0.1,
+        "type": "sine",
+        "volume": 0.5
+      },
+      "success": {
+        "frequency": 1000,
+        "duration": 0.15,
+        "type": "sine",
+        "volume": 0.5
+      },
+      "error": {
+        "frequency": 300,
+        "duration": 0.3,
+        "type": "sawtooth",
+        "volume": 0.5
+      },
+      "notification": {
+        "frequency": 600,
+        "duration": 0.2,
+        "type": "triangle",
+        "volume": 0.4
+      },
+      "listening": {
+        "frequency": 700,
+        "duration": 0.05,
+        "type": "sine",
+        "volume": 0.3
+      }
+    },
+    "confirmations": {
+      "es": ["Listo", "Hecho", "Ok", "Entendido", "Claro"],
+      "en": ["Done", "Got it", "Ok", "Sure", "Alright"],
+      "pt": ["Pronto", "Feito", "Ok", "Entendido", "Certo"]
+    },
+    "errorMessages": {
+      "es": {
+        "notUnderstood": "No te entendi, repite por favor",
+        "lowConfidence": "No estoy seguro, puedes repetir?",
+        "noMicrophone": "No se detecta microfono",
+        "networkError": "Error de conexion"
+      },
+      "en": {
+        "notUnderstood": "I didn't understand, please repeat",
+        "lowConfidence": "I'm not sure, can you repeat?",
+        "noMicrophone": "No microphone detected",
+        "networkError": "Connection error"
+      },
+      "pt": {
+        "notUnderstood": "Nao entendi, repita por favor",
+        "lowConfidence": "Nao tenho certeza, pode repetir?",
+        "noMicrophone": "Microfone nao detectado",
+        "networkError": "Erro de conexao"
+      }
+    }
+  },
+
+  "technicalTerms": {
+    "pronunciations": {
+      "es": {
+        "JavaScript": "yava-script",
+        "TypeScript": "taip-script",
+        "Python": "paizon",
+        "React": "riact",
+        "Vue": "viu",
+        "Angular": "angiular",
+        "Node.js": "noud ye-es",
+        "npm": "en-pe-eme",
+        "git": "guit",
+        "GitHub": "guit-jab",
+        "API": "a-pe-i",
+        "REST": "rest",
+        "GraphQL": "graf-kiu-el",
+        "Docker": "doquer",
+        "Kubernetes": "cubernetis",
+        "AWS": "a-uve-doble-ese",
+        "ELSABRO": "el-sabro",
+        "CLI": "ce-ele-i",
+        "SQL": "ese-cu-ele",
+        "JSON": "yason",
+        "YAML": "yamel"
+      },
+      "en": {
+        "ELSABRO": "el-sah-bro"
+      },
+      "pt": {
+        "JavaScript": "java-script",
+        "ELSABRO": "el-sabro"
+      }
+    },
+    "corrections": {
+      "java script": "JavaScript",
+      "type script": "TypeScript",
+      "pie thon": "Python",
+      "jason": "JSON",
+      "yam el": "YAML",
+      "sequel": "SQL",
+      "post gress": "PostgreSQL",
+      "mongo d b": "MongoDB",
+      "cooper netties": "Kubernetes",
+      "cube control": "kubectl",
+      "git hub": "GitHub",
+      "n p m": "npm",
+      "a w s": "AWS",
+      "a p i": "API",
+      "el sabro": "ELSABRO",
+      "else abro": "ELSABRO"
+    }
+  },
+
+  "performance": {
+    "targets": {
+      "wakeWordLatencyMs": 200,
+      "asrLatencyMs": 500,
+      "intentLatencyMs": 50,
+      "endToEndLatencyMs": 1000,
+      "wakeWordAccuracy": 0.95,
+      "asrWER": 0.10,
+      "intentAccuracy": 0.90
+    },
+    "optimizations": {
+      "preloadModels": true,
+      "warmupOnStart": true,
+      "useWebWorkers": true,
+      "audioProcessingThreads": 2,
+      "batchProcessing": false
+    },
+    "monitoring": {
+      "enabled": true,
+      "metricsEndpoint": "/api/voice/metrics",
+      "logLevel": "info",
+      "logLatency": true,
+      "logAccuracy": true,
+      "sampleRate": 0.1
+    }
+  },
+
+  "privacy": {
+    "audioRetention": "none",
+    "transcriptRetention": "session",
+    "sendToCloud": true,
+    "localProcessingOnly": false,
+    "anonymizeData": true,
+    "consentRequired": true
+  },
+
+  "accessibility": {
+    "screenReaderSupport": true,
+    "visualFeedback": true,
+    "hapticFeedback": true,
+    "highContrastIndicators": true,
+    "voiceSpeedRange": [0.5, 2.0],
+    "keyboardShortcuts": {
+      "toggleVoice": "Ctrl+Shift+V",
+      "pushToTalk": "Space",
+      "cancelCommand": "Escape",
+      "repeatLast": "Ctrl+Shift+R"
+    }
+  },
+
+  "experimental": {
+    "emotionDetection": false,
+    "speakerDiarization": false,
+    "multiSpeaker": false,
+    "sentimentAnalysis": false,
+    "offlineMode": false,
+    "customWakeWordTraining": false
+  }
+}