fix: prevent Ollama indexing from freezing at 69%

- Add dynamic timeout for batch embedding operations based on batch size
- Improve error handling and recovery in batch processing
- Ensure errors are properly thrown to prevent silent failures
- Add immediate error state updates when batch processing fails
- Add better logging for debugging timeout and retry scenarios

Fixes #6849

Author: roomote

src/i18n/locales/en/embeddings.json

@@ -15,7 +15,8 @@
 		"serviceUnavailable": "Ollama service is unavailable (status: {{status}})",
 		"modelNotFound": "Ollama model not found: {{modelId}}",
 		"modelNotEmbeddingCapable": "Ollama model is not embedding capable: {{modelId}}",
-		"hostNotFound": "Ollama host not found: {{baseUrl}}"
+		"hostNotFound": "Ollama host not found: {{baseUrl}}",
+		"batchTimeoutError": "Ollama embedding timed out after {{timeout}} seconds while processing {{count}} texts. Consider reducing batch size or increasing timeout."
 	},
 	"scanner": {
 		"unknownErrorProcessingFile": "Unknown error processing file {{filePath}}",

src/services/code-index/embedders/ollama.ts

@@ -69,8 +69,10 @@ export class CodeIndexOllamaEmbedder implements IEmbedder {
 			// Implementing based on user's specific request structure.
 
 			// Add timeout to prevent indefinite hanging
+			// Use a longer timeout for batch operations as they can take more time
+			const batchTimeout = Math.max(OLLAMA_EMBEDDING_TIMEOUT_MS, texts.length * 2000) // At least 2 seconds per text
 			const controller = new AbortController()
-			const timeoutId = setTimeout(() => controller.abort(), OLLAMA_EMBEDDING_TIMEOUT_MS)
+			const timeoutId = setTimeout(() => controller.abort(), batchTimeout)
 
 			const response = await fetch(url, {
 				method: "POST",
@@ -125,7 +127,15 @@ export class CodeIndexOllamaEmbedder implements IEmbedder {
 
 			// Handle specific error types with better messages
 			if (error.name === "AbortError") {
-				throw new Error(t("embeddings:validation.connectionFailed"))
+				// More specific timeout error message for batch operations
+				const timeoutMessage =
+					texts.length > 1
+						? t("embeddings:ollama.batchTimeoutError", {
+								count: texts.length,
+								timeout: Math.round(Math.max(OLLAMA_EMBEDDING_TIMEOUT_MS, texts.length * 2000) / 1000),
+							})
+						: t("embeddings:validation.connectionFailed")
+				throw new Error(timeoutMessage)
 			} else if (error.message?.includes("fetch failed") || error.code === "ECONNREFUSED") {
 				throw new Error(t("embeddings:ollama.serviceNotRunning", { baseUrl: this.baseUrl }))
 			} else if (error.code === "ENOTFOUND") {

src/services/code-index/orchestrator.ts

@@ -138,11 +138,13 @@ export class CodeIndexOrchestrator {
 
 			const handleFileParsed = (fileBlockCount: number) => {
 				cumulativeBlocksFoundSoFar += fileBlockCount
+				// Update progress immediately when blocks are found
 				this.stateManager.reportBlockIndexingProgress(cumulativeBlocksIndexed, cumulativeBlocksFoundSoFar)
 			}
 
 			const handleBlocksIndexed = (indexedCount: number) => {
 				cumulativeBlocksIndexed += indexedCount
+				// Update progress immediately when blocks are indexed
 				this.stateManager.reportBlockIndexingProgress(cumulativeBlocksIndexed, cumulativeBlocksFoundSoFar)
 			}
 
@@ -154,6 +156,8 @@ export class CodeIndexOrchestrator {
 						batchError,
 					)
 					batchErrors.push(batchError)
+					// Update state to show error immediately
+					this.stateManager.setSystemState("Error", `Indexing error: ${batchError.message}`)
 				},
 				handleBlocksIndexed,
 				handleFileParsed,

src/services/code-index/processors/scanner.ts

@@ -385,8 +385,29 @@ export class DirectoryScanner implements IDirectoryScanner {
 				}
 				// --- End Deletion Step ---
 
-				// Create embeddings for batch
-				const { embeddings } = await this.embedder.createEmbeddings(batchTexts)
+				// Create embeddings for batch with better error context
+				let embeddings: number[][]
+				try {
+					const response = await this.embedder.createEmbeddings(batchTexts)
+					embeddings = response.embeddings
+				} catch (embeddingError: any) {
+					// Log specific details about the embedding failure
+					console.error(
+						`[DirectoryScanner] Embedding creation failed for batch of ${batchTexts.length} texts:`,
+						embeddingError,
+					)
+
+					// Check if it's a timeout error and provide more context
+					if (embeddingError.message?.includes("timed out") || embeddingError.message?.includes("timeout")) {
+						throw new Error(
+							`Embedding timeout for batch of ${batchTexts.length} texts. This may indicate the Ollama service is overloaded or the model is too slow. ${embeddingError.message}`,
+							{ cause: embeddingError },
+						)
+					}
+
+					// Re-throw with additional context
+					throw embeddingError
+				}
 
 				// Prepare points for Qdrant
 				const points = batchBlocks.map((block, index) => {
@@ -420,7 +441,7 @@ export class DirectoryScanner implements IDirectoryScanner {
 			} catch (error) {
 				lastError = error as Error
 				console.error(
-					`[DirectoryScanner] Error processing batch (attempt ${attempts}) in workspace ${scanWorkspace}:`,
+					`[DirectoryScanner] Error processing batch (attempt ${attempts}/${MAX_BATCH_RETRIES}) in workspace ${scanWorkspace}:`,
 					error,
 				)
 				TelemetryService.instance.captureEvent(TelemetryEventName.CODE_INDEX_ERROR, {
@@ -433,6 +454,7 @@ export class DirectoryScanner implements IDirectoryScanner {
 
 				if (attempts < MAX_BATCH_RETRIES) {
 					const delay = INITIAL_RETRY_DELAY_MS * Math.pow(2, attempts - 1)
+					console.log(`[DirectoryScanner] Retrying batch processing in ${delay}ms...`)
 					await new Promise((resolve) => setTimeout(resolve, delay))
 				}
 			}
@@ -454,6 +476,9 @@ export class DirectoryScanner implements IDirectoryScanner {
 					),
 				)
 			}
+			// CRITICAL: Throw the error to ensure it's caught by the orchestrator
+			// This prevents the indexing from appearing to succeed when it actually failed
+			throw lastError
 		}
 	}
 }