One embedding at a time

package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@ztimson/ai-utils",
-	"version": "0.7.1",
+	"version": "0.7.3",
 	"description": "AI Utility library",
 	"author": "Zak Timson",
 	"license": "MIT",

src/llm.ts

@@ -255,11 +255,11 @@ class LLM {
 	/**
 	 * Create a vector representation of a string
 	 * @param {object | string} target Item that will be embedded (objects get converted)
-	 * @param {number} maxTokens Chunking size. More = better context, less = more specific (Search by paragraphs or lines)
+	 * @param {maxTokens?: number, overlapTokens?: number} opts Options for embedding such as chunk sizes
-	 * @param {number} overlapTokens Includes previous X tokens to provide continuity to AI (In addition to max tokens)
 	 * @returns {Promise<Awaited<{index: number, embedding: number[], text: string, tokens: number}>[]>} Chunked embeddings
 	 */
-	embedding(target: object | string, maxTokens = 500, overlapTokens = 50) {
+	async embedding(target: object | string, opts: {maxTokens?: number, overlapTokens?: number} = {}) {
+		let {maxTokens = 500, overlapTokens = 50} = opts;
 		const embed = (text: string): Promise<number[]> => {
 			return new Promise((resolve, reject) => {
 				const worker = new Worker(join(dirname(fileURLToPath(import.meta.url)), 'embedder.js'));
@@ -279,13 +279,13 @@ class LLM {
 				worker.postMessage({text, model: this.ai.options?.embedder || 'bge-small-en-v1.5', modelDir: this.ai.options.path});
 			});
 		};
-		const chunks = this.chunk(target, maxTokens, overlapTokens);
+		const chunks = this.chunk(target, maxTokens, overlapTokens), results: any[] = [];
-		return Promise.all(chunks.map(async (text, index) => ({
+		for(let i = 0; i < chunks.length; i++) {
-			index,
+			const text= chunks[i];
-			embedding: await embed(text),
+			const embedding = await embed(text);
-			text,
+			results.push({index: i, embedding, text, tokens: this.estimateTokens(text)});
-			tokens: this.estimateTokens(text),
+		}
-		})));
+		return results;
 	}
 
 	/**