lumiverse-spindle-types 0.6.38 → 0.6.39

package/dist/dom.d.ts

@@ -1,3 +1,4 @@
+import type { SpindleSTTAPI } from "./speech-to-text.js";
 import type { ChatDTO, ChatMessageDTO, ChatUpdateDTO, ConnectionProfileDTO, LlmMessageDTO, PromptBlockDTO, PromptVariableValuesDTO, RequestInitDTO, TokenCountOptionsDTO, TokenCountResultDTO, WorldBookCreateDTO, WorldBookDTO, WorldBookEntryCreateDTO, WorldBookEntryDTO, WorldBookEntryUpdateDTO, WorldBookUpdateDTO } from "./api.js";
 import type { SpindleComponentsHelper } from "./components.js";
 import type { SpindleHostDescriptorV1, SpindleHostLocaleAPI } from "./host.js";
@@ -1033,6 +1034,8 @@ export interface SpindleRecentChatsPage<TRow = SpindleRecentChat> {
 }
 /** Context object provided to frontend extension modules */
 export interface SpindleFrontendContext {
+    /** Host-managed STT. Feature-detect speech-to-text-v1; unavailable on older hosts. */
+    readonly stt?: SpindleSTTAPI;
     /** Per-document routing identity, available with frontend-session-origin-v1. */
     readonly frontendSessionId?: string;
     /** Immutable host compatibility descriptor for this extension runtime. */

package/dist/index.d.ts

@@ -23,3 +23,5 @@ export type { EntityTypeDTO, EntityStatusDTO, MentionRoleDTO, RelationTypeDTO, R
 export type { SpindleAPI, SpindleGenerateAPI, SpindleConnectionsAPI, SpindlePromptRegex, FrontendProcessHandle, BackendProcessHandle, SpindleBackendProcessContext, SpindleBackendProcessModule, SpindleEmbeddingDriver, SpindleTtsEngine, SpindleSttEngine, SpindleSidecarEndpoint, } from "./spindle-api.js";
 export type { CouncilMember, CouncilMemberContext, SidecarConfig, CouncilSidecarConfig, CouncilToolsSettings, CouncilSettings, CouncilToolResult, CouncilExecutionResult, CachedCouncilResult, CouncilToolCategory, CouncilToolExecution, CouncilToolDefinition, } from "./council.js";
 export { SIDECAR_DEFAULTS, COUNCIL_SIDECAR_DEFAULTS, COUNCIL_TOOLS_DEFAULTS, COUNCIL_SETTINGS_DEFAULTS, } from "./council.js";
+export type { SpindleSTTProvider, SpindleSTTProviderOption, SpindleSTTStatus, SpindleSTTResult, SpindleSTTAudioFrame, SpindleSTTTranscript, SpindleSTTPcmAudio, SpindleSTTOptions, SpindleSTTStartOptions, SpindleSTTSession, SpindleSTTAPI, } from "./speech-to-text.js";
+export { SPINDLE_STT_HOST_CAPABILITIES } from "./speech-to-text.js";

package/dist/index.js

@@ -7,3 +7,4 @@ export { SpindleEvent, CoreEventType } from "./events.js";
 export { ALL_FRONTEND_RUNTIME_CAPABILITIES, isValidFrontendRuntimeCapability, } from "./frontend-capabilities.js";
 export { PERMISSION_DENIED_PREFIX } from "./api.js";
 export { SIDECAR_DEFAULTS, COUNCIL_SIDECAR_DEFAULTS, COUNCIL_TOOLS_DEFAULTS, COUNCIL_SETTINGS_DEFAULTS, } from "./council.js";
+export { SPINDLE_STT_HOST_CAPABILITIES } from "./speech-to-text.js";

package/dist/permissions.d.ts

@@ -18,8 +18,9 @@
  * - "memories"             — CRUD on the Memory Cortex (entities, relations, vaults, chat
  *                            links, consolidations) and long-term chat memory (vectorized
  *                            chat-chunk retrieval, warmup, cache).
- * - "media"                — invoke the backend media pipeline for audio/video conversion,
- *                            transcoding, muxing, and simple image+audio composition.
+ * - "media"                — invoke the host media pipeline for audio/video conversion,
+ *                            transcoding, muxing, and simple image+audio composition; frontend STT
+ *                            and microphone capture through ctx.stt (browser consent also required).
  * - "macro_interceptor"    — transform raw templates before macro parsing/dispatch
  * - "web_search"           — execute searches via the user's configured web search
  *                            provider (e.g. SearXNG) and read the safe view of their

package/dist/speech-to-text.d.ts

@@ -0,0 +1,81 @@
+/** Host-managed speech recognition for frontend Spindle extensions. */
+export type SpindleSTTProvider = 'webspeech' | 'connection' | 'whistle';
+export interface SpindleSTTProviderOption {
+    id: SpindleSTTProvider;
+    name: string;
+    /** Whistle runs entirely on the device. Web Speech behavior depends on the browser. */
+    onDevice: boolean;
+    available: boolean;
+    unavailableReason?: string;
+    supportsAudioTranscription: boolean;
+    languages?: readonly string[];
+}
+export interface SpindleSTTStatus {
+    phase: 'loading' | 'listening' | 'processing';
+    progress?: number;
+}
+/** One engine update: final segments append; interim updates replace the pending segment. */
+export interface SpindleSTTResult {
+    text: string;
+    isFinal: boolean;
+}
+export interface SpindleSTTAudioFrame {
+    amplitude: number;
+    peak: number;
+    frequencies: number[];
+}
+export interface SpindleSTTTranscript {
+    text: string;
+    provider: SpindleSTTProvider;
+    language?: string;
+}
+/** Mono PCM for Whistle. Samples are copied; the caller's buffer is never detached. */
+export interface SpindleSTTPcmAudio {
+    samples: Float32Array;
+    sampleRate: number;
+}
+export interface SpindleSTTOptions {
+    /** Defaults to the user's current Voice & Speech selection. */
+    provider?: SpindleSTTProvider;
+    language?: string;
+    /** Used only with the connection provider; defaults to the user's STT connection. */
+    connectionId?: string;
+    signal?: AbortSignal;
+    onStatus?(status: SpindleSTTStatus): void;
+}
+export interface SpindleSTTStartOptions extends SpindleSTTOptions {
+    /** Defaults to false: finish after confirmed speech followed by silence. */
+    continuous?: boolean;
+    /** Defaults to the user's interim-result preference. */
+    interimResults?: boolean;
+    onResult?(result: SpindleSTTResult): void;
+    onAudioFrame?(frame: SpindleSTTAudioFrame): void;
+}
+export interface SpindleSTTSession {
+    readonly provider: SpindleSTTProvider;
+    /** Resolves when capture starts. Call start() directly from a user gesture. */
+    readonly ready: Promise<void>;
+    /** Resolves after recording ends and the final transcription finishes. */
+    readonly result: Promise<SpindleSTTTranscript>;
+    /** Stops capture and waits for the final transcript; safe to call repeatedly. */
+    stop(): Promise<SpindleSTTTranscript>;
+    /** Releases the microphone immediately; ready/result reject with AbortError if pending. */
+    cancel(): void;
+}
+/** Host-managed frontend STT, available with speech-to-text-v1. No model installation needed. */
+export interface SpindleSTTAPI {
+    /** Capability discovery is free and never requests microphone access. */
+    listProviders(): SpindleSTTProviderOption[];
+    /** Requires media. Prepares Whistle without requesting microphone access. */
+    prepare(options?: SpindleSTTOptions): Promise<void>;
+    /** Requires media and browser microphone permission. One extension capture per document. */
+    start(options?: SpindleSTTStartOptions): SpindleSTTSession;
+    /**
+     * Requires media. Whistle accepts a decodable audio Blob or mono PCM; connection accepts a Blob.
+     * Web Speech cannot transcribe supplied audio. Whistle is capped at ten minutes / 64 MiB.
+     */
+    transcribe(audio: Blob | SpindleSTTPcmAudio, options?: SpindleSTTOptions): Promise<SpindleSTTTranscript>;
+}
+export declare const SPINDLE_STT_HOST_CAPABILITIES: Readonly<{
+    'speech-to-text-v1': 1;
+}>;

package/dist/speech-to-text.js

@@ -0,0 +1 @@
+export const SPINDLE_STT_HOST_CAPABILITIES = Object.freeze({ 'speech-to-text-v1': 1 });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "lumiverse-spindle-types",
-  "version": "0.6.38",
+  "version": "0.6.39",
   "type": "module",
   "types": "./dist/index.d.ts",
   "keywords": [