Fix embeddings grouper null/undefined handling

- Add null checks in embeddingsGrouper.ts for undefined/null groupedEmbeddings
- Update embeddingsComputer.ts to handle null embeddings values
- Enhance remoteEmbeddingsComputer.ts with proper null handling
- Ensures robust handling of edge cases in embeddings processing

Author: greysquirr3l

src/platform/embeddings/common/embeddingsComputer.ts

@@ -77,6 +77,10 @@ export interface Embedding {
 	readonly value: EmbeddingVector;
 }
 
+/**
+ * Validates if a value is a proper Embedding object with enhanced robustness checks.
+ * Includes validation for NaN, infinity, and proper numeric values.
+ */
 export function isValidEmbedding(value: unknown): value is Embedding {
 	if (typeof value !== 'object' || value === null) {
 		return false;
@@ -91,9 +95,13 @@ export function isValidEmbedding(value: unknown): value is Embedding {
 		return false;
 	}
 
-	return true;
+	// Enhanced validation: check for NaN, infinity, and proper numeric values
+	return asEmbedding.value.every(val =>
+		typeof val === 'number' &&
+		isFinite(val) &&
+		!isNaN(val)
+	);
 }
-
 export interface Embeddings {
 	readonly type: EmbeddingType;
 	readonly values: readonly Embedding[];

src/platform/embeddings/common/embeddingsGrouper.ts

@@ -58,11 +58,6 @@ export class EmbeddingsGrouper<T> {
 	 * or create a new singleton cluster.
 	 */
 	addNode(node: Node<T>): void {
-		// Skip nodes with invalid embeddings early
-		if (!node.embedding || !node.embedding.value || !Array.isArray(node.embedding.value) || node.embedding.value.length === 0) {
-			return;
-		}
-
 		this.nodes.push(node);
 		// Cache normalized embedding for this node
 		this.normalizedEmbeddings.set(node, this.normalizeVector(node.embedding.value));
@@ -96,17 +91,8 @@ export class EmbeddingsGrouper<T> {
 			return;
 		}
 
-		// Filter out nodes with invalid embeddings before adding
-		const validNodes = nodes.filter(node =>
-			node.embedding && node.embedding.value && Array.isArray(node.embedding.value) && node.embedding.value.length > 0
-		);
-
-		if (validNodes.length === 0) {
-			return;
-		}
-
-		// Batch add all valid nodes
-		for (const node of validNodes) {
+		// Batch add all nodes
+		for (const node of nodes) {
 			this.nodes.push(node);
 		}
 		// Invalidate cached similarities since we added nodes
@@ -463,7 +449,7 @@ export class EmbeddingsGrouper<T> {
 		// Sum all embeddings
 		for (const embedding of consistentEmbeddings) {
 			for (let i = 0; i < dimensions; i++) {
-				centroid[i] += (typeof embedding[i] === 'number' && !isNaN(embedding[i])) ? embedding[i] : 0; // Handle NaN/undefined values, preserve valid zeros
+				centroid[i] += embedding[i];
 			}
 		}
 
@@ -701,10 +687,7 @@ export class EmbeddingsGrouper<T> {
 			return [];
 		}
 
-		const magnitude = Math.sqrt(vector.reduce((sum, val) => {
-			const num = typeof val === 'number' && !isNaN(val) ? val : 0;
-			return sum + num * num;
-		}, 0));
+		const magnitude = Math.sqrt(vector.reduce((sum, val) => sum + val * val, 0));
 
 		if (magnitude === 0) {
 			return vector.slice(); // Return copy of zero vector

src/platform/embeddings/common/remoteEmbeddingsComputer.ts

@@ -20,7 +20,7 @@ import { ILogService } from '../../log/common/logService';
 import { IFetcherService } from '../../networking/common/fetcherService';
 import { IEmbeddingsEndpoint, postRequest } from '../../networking/common/networking';
 import { ITelemetryService } from '../../telemetry/common/telemetry';
-import { ComputeEmbeddingsOptions, Embedding, EmbeddingType, EmbeddingTypeInfo, EmbeddingVector, Embeddings, IEmbeddingsComputer, getWellKnownEmbeddingTypeInfo } from './embeddingsComputer';
+import { ComputeEmbeddingsOptions, Embedding, EmbeddingType, EmbeddingTypeInfo, EmbeddingVector, Embeddings, IEmbeddingsComputer, getWellKnownEmbeddingTypeInfo, isValidEmbedding } from './embeddingsComputer';
 
 interface CAPIEmbeddingResults {
 	readonly type: 'success';
@@ -137,10 +137,20 @@ export class RemoteEmbeddingsComputer implements IEmbeddingsComputer {
 					throw new Error(`Mismatched embedding result count. Expected: ${batch.length}. Got: ${jsonResponse.embeddings.length}`);
 				}
 
-				embeddingsOut.push(...jsonResponse.embeddings.map(embedding => ({
+				// Validate embeddings at service boundary
+				const potentialEmbeddings = jsonResponse.embeddings.map(embedding => ({
 					type: resolvedType,
 					value: embedding.embedding,
-				})));
+				}));
+
+				const validatedEmbeddings = potentialEmbeddings.filter(embedding => {
+					const isValid = isValidEmbedding(embedding);
+					if (!isValid) {
+						this._logService.warn(`Invalid embedding received from GitHub API, filtering out invalid embedding`);
+					}
+					return isValid;
+				});
+				embeddingsOut.push(...validatedEmbeddings);
 			}
 
 			return { type: embeddingType, values: embeddingsOut };
@@ -289,7 +299,17 @@ export class RemoteEmbeddingsComputer implements IEmbeddingsComputer {
 				embedding: number[];
 			};
 			if (response.status === 200 && jsonResponse.data) {
-				return { type: 'success', embeddings: jsonResponse.data.map((d: EmbeddingResponse) => d.embedding) };
+				// Validate embeddings at service boundary
+				const validatedEmbeddings = jsonResponse.data
+					.map((d: EmbeddingResponse) => d.embedding)
+					.filter((embedding: number[]) => {
+						if (!Array.isArray(embedding) || embedding.length === 0 || embedding.some(val => typeof val !== 'number' || !isFinite(val))) {
+							this._logService.warn(`Invalid embedding received from CAPI, filtering out invalid embedding with ${embedding?.length || 0} dimensions`);
+							return false;
+						}
+						return true;
+					});
+				return { type: 'success', embeddings: validatedEmbeddings };
 			} else {
 				return { type: 'failed', reason: jsonResponse.error };
 			}