fix: Update builder to better work with loops and binary data (no-changelog) (#19040)

packages/@n8n/ai-workflow-builder.ee/evaluations/README.md

@@ -123,6 +123,10 @@ Violations are categorized by severity:
 # Run with default settings
 pnpm eval
 
+# Run a specific test case
+pnpm eval --test-case google-sheets-processing
+pnpm eval --test-case extract-from-file
+
 # With additional generated test cases
 GENERATE_TEST_CASES=true pnpm eval
 

packages/@n8n/ai-workflow-builder.ee/evaluations/chains/test-case-generator.ts

@@ -142,4 +142,16 @@ export const basicTestCases: TestCase[] = [
 		prompt:
 			'Create a multi-agent AI workflow where different AI agents collaborate to research a topic, fact-check information, and compile comprehensive reports.',
 	},
+	{
+		id: 'google-sheets-processing',
+		name: 'Process large Google Sheets data',
+		prompt:
+			'Create a workflow that reads all rows from a Google Sheets document with thousands of customer records. For each row, call an external API to get additional customer data, process the response, and update the row with the enriched information. Handle rate limiting and errors gracefully.',
+	},
+	{
+		id: 'extract-from-file',
+		name: 'Extract data from uploaded files',
+		prompt:
+			'Build a workflow that accepts file uploads through an n8n form. When users upload PDF documents, CSV files, or Excel spreadsheets, automatically extract the text content and data from these files. Transform the extracted data into a structured format and save it to a database or send it via email as a summary.',
+	},
 ];

packages/@n8n/ai-workflow-builder.ee/evaluations/chains/workflow-evaluator.ts

@@ -54,6 +54,7 @@ Evaluate whether the workflow correctly implements what the user EXPLICITLY requ
   - Missing core functionality explicitly requested
   - Incorrect operation logic that prevents the workflow from working
   - Workflows missing a trigger node when they need to start automatically or by some external event
+  - Using Split In Batches node
 - **Major (-15 to -25 points)**:
   - Missing explicitly required data transformations
   - Incomplete implementation of requested features

packages/@n8n/ai-workflow-builder.ee/evaluations/cli/runner.ts

@@ -24,7 +24,7 @@ import { generateMarkdownReport } from '../utils/evaluation-reporter.js';
  * Main CLI evaluation runner that executes all test cases in parallel
  * Supports concurrency control via EVALUATION_CONCURRENCY environment variable
  */
-export async function runCliEvaluation(): Promise<void> {
+export async function runCliEvaluation(testCaseFilter?: string): Promise<void> {
 	console.log(formatHeader('AI Workflow Builder Full Evaluation', 70));
 	console.log();
 	try {
@@ -34,11 +34,24 @@ export async function runCliEvaluation(): Promise<void> {
 		// Determine test cases to run
 		let testCases: TestCase[] = basicTestCases;
 
-		// Optionally generate additional test cases
-		if (shouldGenerateTestCases()) {
-			console.log(pc.blue('➔ Generating additional test cases...'));
-			const generatedCases = await generateTestCases(llm, howManyTestCasesToGenerate());
-			testCases = [...testCases, ...generatedCases];
+		// Filter to single test case if specified
+		if (testCaseFilter) {
+			const filteredCase = testCases.find((tc) => tc.id === testCaseFilter);
+			if (filteredCase) {
+				testCases = [filteredCase];
+				console.log(pc.blue(`➔ Running single test case: ${filteredCase.name}`));
+			} else {
+				console.log(pc.red(`❌ Test case '${testCaseFilter}' not found`));
+				console.log(pc.dim(`Available test cases: ${testCases.map((tc) => tc.id).join(', ')}`));
+				return;
+			}
+		} else {
+			// Optionally generate additional test cases
+			if (shouldGenerateTestCases()) {
+				console.log(pc.blue('➔ Generating additional test cases...'));
+				const generatedCases = await generateTestCases(llm, howManyTestCasesToGenerate());
+				testCases = [...testCases, ...generatedCases];
+			}
 		}
 
 		// Get concurrency from environment

packages/@n8n/ai-workflow-builder.ee/evaluations/index.ts

@@ -14,10 +14,15 @@ export { setupTestEnvironment, createAgent } from './core/environment.js';
 async function main(): Promise<void> {
 	const useLangsmith = process.env.USE_LANGSMITH_EVAL === 'true';
 
+	// Parse command line arguments for single test case
+	const testCaseId = process.argv.includes('--test-case')
+		? process.argv[process.argv.indexOf('--test-case') + 1]
+		: undefined;
+
 	if (useLangsmith) {
 		await runLangsmithEvaluation();
 	} else {
-		await runCliEvaluation();
+		await runCliEvaluation(testCaseId);
 	}
 }
 

packages/@n8n/ai-workflow-builder.ee/src/tools/prompts/main-agent.prompt.ts

@@ -238,6 +238,8 @@ Why: Unconfigured nodes WILL fail at runtime
 For AI-generated structured data, prefer Structured Output Parser nodes over Code nodes.
 Why: Purpose-built parsers are more reliable and handle edge cases better than custom code.
 
+For binary file data, use Extract From File node to extract content from files before processing.
+
 Use Code nodes only for:
 - Simple string manipulations
 - Already structured data (JSON, CSV)
@@ -321,9 +323,10 @@ Anticipate workflow needs and suggest enhancements:
 - Set nodes for data transformation between incompatible formats
 - Schedule Triggers for recurring tasks
 - Error handling for external service calls
-- Split In Batches for large dataset processing
 
 Why: Proactive suggestions create more robust, production-ready workflows
+
+NEVER use Split In Batches nodes.
 </proactive_design>
 
 <parameter_updates>