@matbee/remotemedia-native 0.1.0

package/node-schemas.ts

@@ -0,0 +1,689 @@
+// Auto-generated by RemoteMedia SDK - DO NOT EDIT
+// Run `npm run generate-types` to regenerate
+
+// RuntimeData types (matches Rust enum)
+export type RuntimeDataType = 'audio' | 'video' | 'json' | 'text' | 'binary' | 'tensor' | 'numpy' | 'control' | 'controlmessage';
+
+export interface AudioData {
+  samples: Float32Array;
+  sampleRate: number;
+  channels: number;
+  streamId?: string;
+}
+
+export interface VideoData {
+  pixelData: Uint8Array;
+  width: number;
+  height: number;
+  format: 'yuv420p' | 'rgb24' | 'rgba32' | 'gray8';
+  codec?: 'raw' | 'h264' | 'vp8' | 'vp9' | 'av1';
+  frameNumber?: number;
+  isKeyframe?: boolean;
+}
+
+export interface TensorData {
+  data: Uint8Array;
+  shape: number[];
+  dtype: 'f32' | 'f16' | 'i32' | 'i8' | 'u8';
+}
+
+export interface NumpyArray {
+  data: Uint8Array;
+  shape: number[];
+  dtype: string;
+  strides: number[];
+}
+
+export interface ControlMessage {
+  type: 'start' | 'stop' | 'cancel' | 'flush' | 'config_update' | 'custom';
+  timestamp?: number;
+  segmentId?: string;
+  metadata?: Record<string, unknown>;
+  cancelRange?: { start: number; end: number };
+}
+
+export type RuntimeData =
+  | { type: 'audio'; data: AudioData }
+  | { type: 'video'; data: VideoData }
+  | { type: 'json'; data: Record<string, unknown> }
+  | { type: 'text'; data: string }
+  | { type: 'binary'; data: Uint8Array }
+  | { type: 'tensor'; data: TensorData }
+  | { type: 'numpy'; data: NumpyArray }
+  | { type: 'control'; data: ControlMessage };
+
+/** Performs arithmetic operations on JSON input - Configuration */
+export interface CalculatorNodeConfig {
+  /** Decimal precision for results */
+  precision?: number;
+}
+
+/** Speech-to-text transcription using Whisper - Configuration */
+export interface WhisperNodeConfig {
+  /** Language code (null for auto-detect) */
+  language?: string;
+  /** Whisper model size */
+  model?: 'tiny' | 'base' | 'small' | 'medium' | 'large' | 'large-v3';
+  /** Task type */
+  task?: 'transcribe' | 'translate';
+}
+
+/** Splits audio into fixed-size chunks - Configuration */
+export interface AudioChunkerConfig {
+  /** Chunk duration in milliseconds */
+  chunk_size_ms?: number;
+}
+
+/** Collects text chunks into complete utterances - Configuration */
+export interface TextCollectorConfig {
+  /** Delimiter to split on */
+  delimiter?: string;
+  /** Flush buffer when silence detected */
+  flush_on_silence?: boolean;
+}
+
+/** Voice Activity Detection using Silero VAD model - Configuration */
+export interface SileroVADConfig {
+  /** Minimum silence duration in ms */
+  min_silence_duration_ms?: number;
+  /** Minimum speech duration in ms */
+  min_speech_duration_ms?: number;
+  /** Speech probability threshold */
+  threshold?: number;
+}
+
+/** Resamples audio to target sample rate - Configuration */
+export interface AudioResampleConfig {
+  /** Target sample rate in Hz */
+  target_sample_rate?: number;
+}
+
+/** Text-to-speech synthesis using Kokoro TTS - Configuration */
+export interface KokoroTTSNodeConfig {
+  /** Language code */
+  language?: 'en-us' | 'en-gb' | 'es' | 'fr' | 'de' | 'it' | 'ja' | 'ko' | 'pt-br' | 'zh';
+  /** Speech speed multiplier */
+  speed?: number;
+  /** Voice ID to use */
+  voice?: 'af_bella' | 'af_nicole' | 'af_sarah' | 'af_sky' | 'am_adam' | 'am_michael' | 'bf_emma' | 'bf_isabella' | 'bm_george' | 'bm_lewis';
+}
+
+/** Flips video frames horizontally or vertically - Configuration */
+export interface VideoFlipConfig {
+  /** Flip horizontally */
+  horizontal?: boolean;
+  /** Flip vertically */
+  vertical?: boolean;
+}
+
+/** Node metadata from registry */
+export interface NodeMetadata {
+  nodeType: string;
+  description?: string;
+  category?: string;
+  accepts: RuntimeDataType[];
+  produces: RuntimeDataType[];
+  isPython: boolean;
+  streaming: boolean;
+  multiOutput: boolean;
+}
+
+/** All registered node types */
+export type NodeType =
+  | 'CalculatorNode'
+  | 'WhisperNode'
+  | 'AudioChunker'
+  | 'PassThrough'
+  | 'TextCollector'
+  | 'Echo'
+  | 'SileroVAD'
+  | 'AudioResample'
+  | 'KokoroTTSNode'
+  | 'VideoFlip';
+
+/** Node type to config type mapping */
+export interface NodeConfigMap {
+  'CalculatorNode': CalculatorNodeConfig;
+  'WhisperNode': WhisperNodeConfig;
+  'AudioChunker': AudioChunkerConfig;
+  'PassThrough': Record<string, unknown>;
+  'TextCollector': TextCollectorConfig;
+  'Echo': Record<string, unknown>;
+  'SileroVAD': SileroVADConfig;
+  'AudioResample': AudioResampleConfig;
+  'KokoroTTSNode': KokoroTTSNodeConfig;
+  'VideoFlip': VideoFlipConfig;
+}
+
+/** Pipeline node with typed config */
+export interface PipelineNode<T extends NodeType = NodeType> {
+  id: string;
+  nodeType: T;
+  config?: T extends keyof NodeConfigMap ? NodeConfigMap[T] : Record<string, unknown>;
+}
+
+/** Pipeline manifest */
+export interface PipelineManifest {
+  version: string;
+  metadata?: { name?: string; description?: string; [key: string]: unknown };
+  nodes: PipelineNode[];
+  connections: Array<{
+    source: string;
+    sourcePort?: string;
+    destination: string;
+    destinationPort?: string;
+  }>;
+}
+
+/** All node schemas (for runtime introspection) */
+export const nodeSchemas: NodeMetadata[] = [
+  {
+    "nodeType": "CalculatorNode",
+    "description": "Performs arithmetic operations on JSON input",
+    "category": "utility",
+    "accepts": [
+      "json"
+    ],
+    "produces": [
+      "json"
+    ],
+    "isPython": false,
+    "streaming": true,
+    "multiOutput": false
+  },
+  {
+    "nodeType": "WhisperNode",
+    "description": "Speech-to-text transcription using Whisper",
+    "category": "ml",
+    "accepts": [
+      "audio"
+    ],
+    "produces": [
+      "text",
+      "json"
+    ],
+    "isPython": true,
+    "streaming": true,
+    "multiOutput": false
+  },
+  {
+    "nodeType": "AudioChunker",
+    "description": "Splits audio into fixed-size chunks",
+    "category": "audio",
+    "accepts": [
+      "audio"
+    ],
+    "produces": [
+      "audio"
+    ],
+    "isPython": false,
+    "streaming": true,
+    "multiOutput": true
+  },
+  {
+    "nodeType": "PassThrough",
+    "description": "Passes input through unchanged",
+    "category": "utility",
+    "accepts": [
+      "audio",
+      "video",
+      "json",
+      "text",
+      "binary",
+      "tensor",
+      "numpy",
+      "controlmessage"
+    ],
+    "produces": [
+      "audio",
+      "video",
+      "json",
+      "text",
+      "binary",
+      "tensor",
+      "numpy",
+      "controlmessage"
+    ],
+    "isPython": false,
+    "streaming": true,
+    "multiOutput": false
+  },
+  {
+    "nodeType": "TextCollector",
+    "description": "Collects text chunks into complete utterances",
+    "category": "text",
+    "accepts": [
+      "text"
+    ],
+    "produces": [
+      "text"
+    ],
+    "isPython": false,
+    "streaming": true,
+    "multiOutput": false
+  },
+  {
+    "nodeType": "Echo",
+    "description": "Passes input through unchanged (for testing)",
+    "category": "utility",
+    "accepts": [
+      "audio",
+      "video",
+      "json",
+      "text",
+      "binary",
+      "tensor",
+      "numpy",
+      "controlmessage"
+    ],
+    "produces": [
+      "audio",
+      "video",
+      "json",
+      "text",
+      "binary",
+      "tensor",
+      "numpy",
+      "controlmessage"
+    ],
+    "isPython": false,
+    "streaming": true,
+    "multiOutput": false
+  },
+  {
+    "nodeType": "SileroVAD",
+    "description": "Voice Activity Detection using Silero VAD model",
+    "category": "audio",
+    "accepts": [
+      "audio"
+    ],
+    "produces": [
+      "audio",
+      "controlmessage"
+    ],
+    "isPython": false,
+    "streaming": true,
+    "multiOutput": false
+  },
+  {
+    "nodeType": "AudioResample",
+    "description": "Resamples audio to target sample rate",
+    "category": "audio",
+    "accepts": [
+      "audio"
+    ],
+    "produces": [
+      "audio"
+    ],
+    "isPython": false,
+    "streaming": true,
+    "multiOutput": false
+  },
+  {
+    "nodeType": "KokoroTTSNode",
+    "description": "Text-to-speech synthesis using Kokoro TTS",
+    "category": "ml",
+    "accepts": [
+      "text"
+    ],
+    "produces": [
+      "audio"
+    ],
+    "isPython": true,
+    "streaming": true,
+    "multiOutput": true
+  },
+  {
+    "nodeType": "VideoFlip",
+    "description": "Flips video frames horizontally or vertically",
+    "category": "video",
+    "accepts": [
+      "video"
+    ],
+    "produces": [
+      "video"
+    ],
+    "isPython": false,
+    "streaming": true,
+    "multiOutput": false
+  }
+];
+
+// =============================================================================
+// Node Builder Classes
+// =============================================================================
+
+/**
+ * Base class for all node builders.
+ * Provides type-safe construction of pipeline nodes.
+ */
+export abstract class NodeBuilder<T extends NodeType = NodeType, C = unknown> {
+  readonly id: string;
+  readonly nodeType: T;
+  readonly config?: C;
+
+  constructor(id: string, nodeType: T, config?: C) {
+    this.id = id;
+    this.nodeType = nodeType;
+    this.config = config;
+  }
+
+  /** Convert to PipelineNode format for manifest */
+  toPipelineNode(): PipelineNode<T> {
+    return {
+      id: this.id,
+      nodeType: this.nodeType,
+      config: this.config as T extends keyof NodeConfigMap ? NodeConfigMap[T] : Record<string, unknown>,
+    };
+  }
+
+  /** Create connection to another node */
+  connectTo(target: NodeBuilder<NodeType, unknown> | string, sourcePort?: string, destinationPort?: string): PipelineConnection {
+    const targetId = typeof target === 'string' ? target : target.id;
+    return {
+      source: this.id,
+      sourcePort,
+      destination: targetId,
+      destinationPort,
+    };
+  }
+}
+
+/** Connection between pipeline nodes */
+export interface PipelineConnection {
+  source: string;
+  sourcePort?: string;
+  destination: string;
+  destinationPort?: string;
+}
+
+/**
+ * Performs arithmetic operations on JSON input
+ *
+ * @example
+ * ```typescript
+ * const node = new CalculatorNode('my-calculatornode', { });
+ * pipeline.addNode(node);
+ * ```
+ */
+export class CalculatorNode extends NodeBuilder<'CalculatorNode', CalculatorNodeConfig> {
+  static readonly nodeType = 'CalculatorNode' as const;
+  static readonly accepts: RuntimeDataType[] = ['json'];
+  static readonly produces: RuntimeDataType[] = ['json'];
+
+  constructor(id: string, config?: CalculatorNodeConfig) {
+    super(id, 'CalculatorNode', config);
+  }
+}
+
+/**
+ * Speech-to-text transcription using Whisper
+ *
+ * @example
+ * ```typescript
+ * const node = new WhisperNode('my-whispernode', { });
+ * pipeline.addNode(node);
+ * ```
+ */
+export class WhisperNode extends NodeBuilder<'WhisperNode', WhisperNodeConfig> {
+  static readonly nodeType = 'WhisperNode' as const;
+  static readonly accepts: RuntimeDataType[] = ['audio'];
+  static readonly produces: RuntimeDataType[] = ['text', 'json'];
+
+  constructor(id: string, config?: WhisperNodeConfig) {
+    super(id, 'WhisperNode', config);
+  }
+}
+
+/**
+ * Splits audio into fixed-size chunks
+ *
+ * @example
+ * ```typescript
+ * const node = new AudioChunker('my-audiochunker', { });
+ * pipeline.addNode(node);
+ * ```
+ */
+export class AudioChunker extends NodeBuilder<'AudioChunker', AudioChunkerConfig> {
+  static readonly nodeType = 'AudioChunker' as const;
+  static readonly accepts: RuntimeDataType[] = ['audio'];
+  static readonly produces: RuntimeDataType[] = ['audio'];
+
+  constructor(id: string, config?: AudioChunkerConfig) {
+    super(id, 'AudioChunker', config);
+  }
+}
+
+/**
+ * Passes input through unchanged
+ *
+ * @example
+ * ```typescript
+ * const node = new PassThrough('my-passthrough');
+ * pipeline.addNode(node);
+ * ```
+ */
+export class PassThrough extends NodeBuilder<'PassThrough', Record<string, unknown>> {
+  static readonly nodeType = 'PassThrough' as const;
+  static readonly accepts: RuntimeDataType[] = ['audio', 'video', 'json', 'text', 'binary', 'tensor', 'numpy', 'controlmessage'];
+  static readonly produces: RuntimeDataType[] = ['audio', 'video', 'json', 'text', 'binary', 'tensor', 'numpy', 'controlmessage'];
+
+  constructor(id: string, config?: Record<string, unknown>) {
+    super(id, 'PassThrough', config);
+  }
+}
+
+/**
+ * Collects text chunks into complete utterances
+ *
+ * @example
+ * ```typescript
+ * const node = new TextCollector('my-textcollector', { });
+ * pipeline.addNode(node);
+ * ```
+ */
+export class TextCollector extends NodeBuilder<'TextCollector', TextCollectorConfig> {
+  static readonly nodeType = 'TextCollector' as const;
+  static readonly accepts: RuntimeDataType[] = ['text'];
+  static readonly produces: RuntimeDataType[] = ['text'];
+
+  constructor(id: string, config?: TextCollectorConfig) {
+    super(id, 'TextCollector', config);
+  }
+}
+
+/**
+ * Passes input through unchanged (for testing)
+ *
+ * @example
+ * ```typescript
+ * const node = new Echo('my-echo');
+ * pipeline.addNode(node);
+ * ```
+ */
+export class Echo extends NodeBuilder<'Echo', Record<string, unknown>> {
+  static readonly nodeType = 'Echo' as const;
+  static readonly accepts: RuntimeDataType[] = ['audio', 'video', 'json', 'text', 'binary', 'tensor', 'numpy', 'controlmessage'];
+  static readonly produces: RuntimeDataType[] = ['audio', 'video', 'json', 'text', 'binary', 'tensor', 'numpy', 'controlmessage'];
+
+  constructor(id: string, config?: Record<string, unknown>) {
+    super(id, 'Echo', config);
+  }
+}
+
+/**
+ * Voice Activity Detection using Silero VAD model
+ *
+ * @example
+ * ```typescript
+ * const node = new SileroVAD('my-silerovad', { });
+ * pipeline.addNode(node);
+ * ```
+ */
+export class SileroVAD extends NodeBuilder<'SileroVAD', SileroVADConfig> {
+  static readonly nodeType = 'SileroVAD' as const;
+  static readonly accepts: RuntimeDataType[] = ['audio'];
+  static readonly produces: RuntimeDataType[] = ['audio', 'controlmessage'];
+
+  constructor(id: string, config?: SileroVADConfig) {
+    super(id, 'SileroVAD', config);
+  }
+}
+
+/**
+ * Resamples audio to target sample rate
+ *
+ * @example
+ * ```typescript
+ * const node = new AudioResample('my-audioresample', { });
+ * pipeline.addNode(node);
+ * ```
+ */
+export class AudioResample extends NodeBuilder<'AudioResample', AudioResampleConfig> {
+  static readonly nodeType = 'AudioResample' as const;
+  static readonly accepts: RuntimeDataType[] = ['audio'];
+  static readonly produces: RuntimeDataType[] = ['audio'];
+
+  constructor(id: string, config?: AudioResampleConfig) {
+    super(id, 'AudioResample', config);
+  }
+}
+
+/**
+ * Text-to-speech synthesis using Kokoro TTS
+ *
+ * @example
+ * ```typescript
+ * const node = new KokoroTTSNode('my-kokorottsnode', { });
+ * pipeline.addNode(node);
+ * ```
+ */
+export class KokoroTTSNode extends NodeBuilder<'KokoroTTSNode', KokoroTTSNodeConfig> {
+  static readonly nodeType = 'KokoroTTSNode' as const;
+  static readonly accepts: RuntimeDataType[] = ['text'];
+  static readonly produces: RuntimeDataType[] = ['audio'];
+
+  constructor(id: string, config?: KokoroTTSNodeConfig) {
+    super(id, 'KokoroTTSNode', config);
+  }
+}
+
+/**
+ * Flips video frames horizontally or vertically
+ *
+ * @example
+ * ```typescript
+ * const node = new VideoFlip('my-videoflip', { });
+ * pipeline.addNode(node);
+ * ```
+ */
+export class VideoFlip extends NodeBuilder<'VideoFlip', VideoFlipConfig> {
+  static readonly nodeType = 'VideoFlip' as const;
+  static readonly accepts: RuntimeDataType[] = ['video'];
+  static readonly produces: RuntimeDataType[] = ['video'];
+
+  constructor(id: string, config?: VideoFlipConfig) {
+    super(id, 'VideoFlip', config);
+  }
+}
+
+/** Namespace containing all node builder classes */
+export const Nodes = {
+  CalculatorNode,
+  WhisperNode,
+  AudioChunker,
+  PassThrough,
+  TextCollector,
+  Echo,
+  SileroVAD,
+  AudioResample,
+  KokoroTTSNode,
+  VideoFlip
+} as const;
+
+/**
+ * Fluent pipeline builder for constructing manifests.
+ *
+ * @example
+ * ```typescript
+ * const manifest = new PipelineBuilder('1.0')
+ *   .name('My Pipeline')
+ *   .add(new SileroVAD('vad', { threshold: 0.6 }))
+ *   .add(new WhisperNode('whisper', { model: 'base' }))
+ *   .connect('vad', 'whisper')
+ *   .build();
+ * ```
+ */
+export class PipelineBuilder {
+  private version: string;
+  private metadata: { name?: string; description?: string; [key: string]: unknown } = {};
+  private nodes: PipelineNode[] = [];
+  private connections: PipelineConnection[] = [];
+
+  constructor(version: string = '1.0') {
+    this.version = version;
+  }
+
+  /** Set pipeline name */
+  name(name: string): this {
+    this.metadata.name = name;
+    return this;
+  }
+
+  /** Set pipeline description */
+  description(description: string): this {
+    this.metadata.description = description;
+    return this;
+  }
+
+  /** Add metadata key-value */
+  meta(key: string, value: unknown): this {
+    this.metadata[key] = value;
+    return this;
+  }
+
+  /** Add a node to the pipeline */
+  add(node: NodeBuilder<NodeType, unknown>): this {
+    this.nodes.push(node.toPipelineNode());
+    return this;
+  }
+
+  /** Add a raw node definition */
+  addRaw<T extends NodeType>(node: PipelineNode<T>): this {
+    this.nodes.push(node);
+    return this;
+  }
+
+  /** Connect two nodes */
+  connect(
+    source: NodeBuilder<NodeType, unknown> | string,
+    destination: NodeBuilder<NodeType, unknown> | string,
+    sourcePort?: string,
+    destinationPort?: string
+  ): this {
+    const sourceId = typeof source === 'string' ? source : source.id;
+    const destId = typeof destination === 'string' ? destination : destination.id;
+    this.connections.push({
+      source: sourceId,
+      sourcePort,
+      destination: destId,
+      destinationPort,
+    });
+    return this;
+  }
+
+  /** Build the final pipeline manifest */
+  build(): PipelineManifest {
+    return {
+      version: this.version,
+      metadata: Object.keys(this.metadata).length > 0 ? this.metadata : undefined,
+      nodes: this.nodes,
+      connections: this.connections,
+    };
+  }
+
+  /** Convert to JSON string */
+  toJson(): string {
+    return JSON.stringify(this.build(), null, 2);
+  }
+}