pr comments

src/VectorStore.ts

@@ -303,7 +303,7 @@ export class VectorStore {
 		return new Map(dbFile.embeddings)
 	}
 
-	private similarityBatch(searchVectors: Vector[], possibleMatch: Vector): number {
+	private computeSimilarity(searchVectors: Vector[], possibleMatch: Vector): number {
 		let highestSimilarity = similarity(searchVectors.first(), possibleMatch)
 		for (const searchVector of searchVectors) {
 			const currentSimilarity = similarity(searchVector, possibleMatch)
@@ -314,14 +314,14 @@ export class VectorStore {
 		return highestSimilarity
 	}
 
-	getNearestVectorsBatch(searchVectors: Vector[], resultNumber: number, relevanceThreshold: number): NearestVectorResult[] {
+	getNearestVectors(searchVectors: Vector[], resultNumber: number, relevanceThreshold: number): NearestVectorResult[] {
 		const nearestVectors: NearestVectorResult[] = []
 
 		for (const entry of this.embeddings.entries()) {
 			const filePath = entry[0]
 			const fileEntry = entry[1]
 			if (fileEntry.embedding && fileEntry.embedding.length) {
-				const fileSimilarity = this.similarityBatch(searchVectors, fileEntry.embedding)
+				const fileSimilarity = this.computeSimilarity(searchVectors, fileEntry.embedding)
 				nearestVectors.push({
 					path: filePath,
 					chunk: undefined,
@@ -330,7 +330,7 @@ export class VectorStore {
 			}
 			fileEntry.chunks.forEach(chunk => {
 				if (chunk.embedding && chunk.embedding.length) {
-					const chunkSimilarity = this.similarityBatch(searchVectors, chunk.embedding)
+					const chunkSimilarity = this.computeSimilarity(searchVectors, chunk.embedding)
 					nearestVectors.push({
 						path: filePath,
 						chunk: chunk.contents,

src/main.ts

@@ -138,7 +138,7 @@ export default class VaultChat extends Plugin {
 	}
 
 	async askChatGpt(question: string) {
-		// HyDE: request note that answers the question
+		// HyDE: request note that answers the question https://github.com/texttron/hyde
 		const conversation: Array<ChatCompletionRequestMessage> = []
 		const hydeMessage: ChatCompletionRequestMessage = {
 			role: ChatCompletionResponseMessageRoleEnum.User,
@@ -154,18 +154,18 @@ export default class VaultChat extends Plugin {
 		// create embeddings for that note and all of its blocks
 		const hydeNote = hydeResponse.choices[0].message.content
 		const hydeNoteBlocks = parseMarkdown(hydeNote, '')
-		const hydeBlocksStrings = hydeNoteBlocks.map(t => `${t.path} ${t.localHeading} ${t.content}`)
-		hydeBlocksStrings.push(hydeNote)
-		hydeBlocksStrings.push(question) // include the users raw question
-		const embeddingsResponse = await this.openAIHandler.createEmbeddingBatch(hydeBlocksStrings)
+		const queryBlockStrings = hydeNoteBlocks.map(t => `${t.path} ${t.localHeading} ${t.content}`)
+		queryBlockStrings.push(hydeNote)
+		queryBlockStrings.push(question) // include the users raw question
+		const embeddingsResponse = await this.openAIHandler.createEmbeddingBatch(queryBlockStrings)
 		const embeddings = embeddingsResponse?.data
 		if (!embeddings) {
 			console.error(`Failed to get embeddings for hyde response`)
 			return
 		}
 		const searchVectors = embeddings.map(e => e.embedding)
 		// search for matches
-		const nearestVectors = this.vectorStore.getNearestVectorsBatch(searchVectors, 8, this.settings.relevanceThreshold)
+		const nearestVectors = this.vectorStore.getNearestVectors(searchVectors, 8, this.settings.relevanceThreshold)
 		const results = await Promise.all(nearestVectors.map(async (nearest, i) => {
 			let name = nearest.path.split('/').last() || ''
 			let contents = nearest.chunk