Auto download teseract models

package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@ztimson/ai-utils",
-	"version": "0.2.0",
+	"version": "0.2.3",
 	"description": "AI Utility library",
 	"author": "Zak Timson",
 	"license": "MIT",

src/ai.ts

@@ -8,10 +8,10 @@ export type AiOptions = LLMOptions & {
 		binary: string;
 		/** Model: `ggml-base.en.bin` */
 		model: string;
+	}
 	/** Path to models */
 	path: string;
 }
-}
 
 export class Ai {
 	private downloads: {[key: string]: Promise<string>} = {};
@@ -25,6 +25,7 @@ export class Ai {
 	vision!: Vision;
 
 	constructor(public readonly options: AiOptions) {
+		process.env.TRANSFORMERS_CACHE = options.path;
 		this.audio = new Audio(this);
 		this.language = new LLM(this);
 		this.vision = new Vision(this);

src/audio.ts

@@ -48,7 +48,7 @@ export class Audio {
 	async downloadAsrModel(model: string = this.whisperModel): Promise<string> {
 		if(!this.ai.options.whisper?.binary) throw new Error('Whisper not configured');
 		if(!model.endsWith('.bin')) model += '.bin';
-		const p = Path.join(this.ai.options.whisper.path, model);
+		const p = Path.join(this.ai.options.path, model);
 		if(await fs.stat(p).then(() => true).catch(() => false)) return p;
 		if(!!this.downloads[model]) return this.downloads[model];
 		this.downloads[model] = fetch(`https://huggingface.co/ggerganov/whisper.cpp/resolve/main/${model}`)

src/llm.ts

@@ -136,6 +136,18 @@ export class LLM {
 		return [{role: 'assistant', content: `Conversation Summary: ${summary}`, timestamp: Date.now()}, ...recent];
 	}
 
+	cosineSimilarity(v1: number[], v2: number[]): number {
+		if (v1.length !== v2.length) throw new Error('Vectors must be same length');
+		let dotProduct = 0, normA = 0, normB = 0;
+		for (let i = 0; i < v1.length; i++) {
+			dotProduct += v1[i] * v2[i];
+			normA += v1[i] * v1[i];
+			normB += v2[i] * v2[i];
+		}
+		const denominator = Math.sqrt(normA) * Math.sqrt(normB);
+		return denominator === 0 ? 0 : dotProduct / denominator;
+	}
+
 	embedding(target: object | string, maxTokens = 500, overlapTokens = 50) {
 		const objString = (obj: any, path = ''): string[] => {
 			if(obj === null || obj === undefined) return [];
@@ -205,24 +217,12 @@ export class LLM {
 	 */
 	fuzzyMatch(target: string, ...searchTerms: string[]) {
 		if(searchTerms.length < 2) throw new Error('Requires at least 2 strings to compare');
-
 		const vector = (text: string, dimensions: number = 10): number[] => {
 			return text.toLowerCase().split('').map((char, index) =>
 				(char.charCodeAt(0) * (index + 1)) % dimensions / dimensions).slice(0, dimensions);
 		}
-
-		const cosineSimilarity = (v1: number[], v2: number[]): number => {
-			if (v1.length !== v2.length) throw new Error('Vectors must be same length');
-			const tensor1 = tf.tensor1d(v1), tensor2 = tf.tensor1d(v2)
-			const dotProduct = tf.dot(tensor1, tensor2)
-			const magnitude1 = tf.norm(tensor1)
-			const magnitude2 = tf.norm(tensor2)
-			if(magnitude1.dataSync()[0] === 0 || magnitude2.dataSync()[0] === 0) return 0
-			return dotProduct.dataSync()[0] / (magnitude1.dataSync()[0] * magnitude2.dataSync()[0])
-		}
-
 		const v = vector(target);
-		const similarities = searchTerms.map(t => vector(t)).map(refVector => cosineSimilarity(v, refVector))
+		const similarities = searchTerms.map(t => vector(t)).map(refVector => this.cosineSimilarity(v, refVector))
 		return {avg: similarities.reduce((acc, s) => acc + s, 0) / similarities.length, max: Math.max(...similarities), similarities}
 	}
 

src/vision.ts

@@ -15,7 +15,7 @@ export class Vision {
 		return {
 			abort: () => { worker?.terminate(); },
 			response: new Promise(async res => {
-				worker = await createWorker('eng');
+				worker = await createWorker('eng', 1, {cachePath: this.ai.options.path});
 				const {data} = await worker.recognize(path);
 				await worker.terminate();
 				res(data.text.trim() || null);