118 lines
4.2 KiB
TypeScript
118 lines
4.2 KiB
TypeScript
/**
|
|
* WASM bindings for OsPipe - use in browser-based pipes.
|
|
*
|
|
* This module provides a thin wrapper around the @ruvector/ospipe-wasm package,
|
|
* exposing vector search, embedding, deduplication, and safety checking
|
|
* capabilities that run entirely client-side via WebAssembly.
|
|
*
|
|
* @packageDocumentation
|
|
*/
|
|
/** A single search result from the WASM vector index. */
|
|
export interface WasmSearchResult {
|
|
/** Unique identifier for the indexed entry */
|
|
id: string;
|
|
/** Similarity score (higher is more similar) */
|
|
score: number;
|
|
/** JSON-encoded metadata string */
|
|
metadata: string;
|
|
}
|
|
/** Configuration options for WASM initialization. */
|
|
export interface OsPipeWasmOptions {
|
|
/** Embedding vector dimension (default: 384) */
|
|
dimension?: number;
|
|
}
|
|
/** The initialized WASM instance interface. */
|
|
export interface OsPipeWasmInstance {
|
|
/**
|
|
* Insert a vector into the index.
|
|
*
|
|
* @param id - Unique identifier for the entry
|
|
* @param embedding - Float32Array embedding vector
|
|
* @param metadata - JSON-encoded metadata string
|
|
* @param timestamp - Unix timestamp in milliseconds (default: Date.now())
|
|
*/
|
|
insert(id: string, embedding: Float32Array, metadata: string, timestamp?: number): void;
|
|
/**
|
|
* Search for the k nearest neighbors to the query embedding.
|
|
*
|
|
* @param queryEmbedding - Float32Array query vector
|
|
* @param k - Number of results to return (default: 10)
|
|
* @returns Array of search results ranked by similarity
|
|
*/
|
|
search(queryEmbedding: Float32Array, k?: number): WasmSearchResult[];
|
|
/**
|
|
* Search with a time range filter applied before ranking.
|
|
*
|
|
* @param queryEmbedding - Float32Array query vector
|
|
* @param k - Number of results to return
|
|
* @param startTime - Start of time range (Unix ms)
|
|
* @param endTime - End of time range (Unix ms)
|
|
* @returns Array of filtered search results
|
|
*/
|
|
searchFiltered(queryEmbedding: Float32Array, k: number, startTime: number, endTime: number): WasmSearchResult[];
|
|
/**
|
|
* Check if an embedding is a near-duplicate of an existing entry.
|
|
*
|
|
* @param embedding - Float32Array embedding to check
|
|
* @param threshold - Similarity threshold 0-1 (default: 0.95)
|
|
* @returns True if a duplicate is found above the threshold
|
|
*/
|
|
isDuplicate(embedding: Float32Array, threshold?: number): boolean;
|
|
/**
|
|
* Generate an embedding vector from text using the built-in ONNX model.
|
|
*
|
|
* @param text - Input text to embed
|
|
* @returns Float32Array embedding vector
|
|
*/
|
|
embedText(text: string): Float32Array;
|
|
/**
|
|
* Run a safety check on content, returning the recommended action.
|
|
*
|
|
* @param content - Content string to check
|
|
* @returns "allow", "redact", or "deny"
|
|
*/
|
|
safetyCheck(content: string): "allow" | "redact" | "deny";
|
|
/**
|
|
* Route a query to the optimal query type.
|
|
*
|
|
* @param query - Natural language query string
|
|
* @returns Recommended query route type
|
|
*/
|
|
routeQuery(query: string): string;
|
|
/** Number of entries currently in the index. */
|
|
readonly size: number;
|
|
/**
|
|
* Get index statistics as a JSON string.
|
|
*
|
|
* @returns JSON-encoded statistics object
|
|
*/
|
|
stats(): string;
|
|
}
|
|
/**
|
|
* Load and initialize the OsPipe WASM module.
|
|
*
|
|
* This function dynamically imports the @ruvector/ospipe-wasm package,
|
|
* initializes the WebAssembly module, and returns a typed wrapper
|
|
* around the raw WASM bindings.
|
|
*
|
|
* @param options - WASM initialization options
|
|
* @returns Initialized WASM instance with typed methods
|
|
* @throws {Error} If the WASM module fails to load or initialize
|
|
*
|
|
* @example
|
|
* ```typescript
|
|
* import { initOsPipeWasm } from "@ruvector/ospipe/wasm";
|
|
*
|
|
* const wasm = await initOsPipeWasm({ dimension: 384 });
|
|
*
|
|
* // Embed and insert
|
|
* const embedding = wasm.embedText("hello world");
|
|
* wasm.insert("doc-1", embedding, JSON.stringify({ app: "test" }));
|
|
*
|
|
* // Search
|
|
* const query = wasm.embedText("greetings");
|
|
* const results = wasm.search(query, 5);
|
|
* ```
|
|
*/
|
|
export declare function initOsPipeWasm(options?: OsPipeWasmOptions): Promise<OsPipeWasmInstance>;
|
|
//# sourceMappingURL=wasm.d.ts.map
|