Ingest: Add post-indexing content date resolution

Directory.Packages.props

@@ -128,5 +128,6 @@
       <IncludeAssets>runtime; build; native; contentfiles; analyzers; buildtransitive</IncludeAssets>
     </PackageVersion>
     <PackageVersion Include="xunit.v3" Version="2.0.2" />
+    <PackageVersion Include="Testcontainers.Elasticsearch" Version="4.6.0" />
   </ItemGroup>
 </Project>

docs-builder.slnx

@@ -88,6 +88,7 @@
 	<Project Path="tests-integration/Elastic.Documentation.Api.IntegrationTests/Elastic.Documentation.Api.IntegrationTests.csproj" />
     <Project Path="tests-integration/Search.IntegrationTests/Search.IntegrationTests.csproj" />
 	  <Project Path="tests-integration/Mcp.Remote.IntegrationTests/Mcp.Remote.IntegrationTests.csproj" />
+	  <Project Path="tests-integration/Elastic.ContentDateEnrichment.IntegrationTests/Elastic.ContentDateEnrichment.IntegrationTests.csproj" />
   </Folder>
   <Folder Name="/tests/">
     <File Path="tests/Directory.Build.props" />

src/Elastic.Markdown/Exporters/Elasticsearch/ContentDateEnrichment.cs

@@ -69,6 +69,23 @@ public async Task SyncLookupIndexAsync(string lexicalAlias, Cancel ct)
 		logger.LogInformation("Content date lookup sync complete");
 	}
 
+	/// <summary>
+	/// Runs _update_by_query with the enrichment pipeline on all documents.
+	/// HashedBulkUpdate uses bulk update actions which skip ingest pipelines,
+	/// so the pipeline never fires during normal indexing. This post-indexing step
+	/// applies the pipeline to resolve content_last_updated for every document:
+	/// the pipeline compares each document's content_hash against the lookup from
+	/// the previous run and either preserves the old date or stamps a new one.
+	/// </summary>
+	public async Task ResolveContentDatesAsync(string indexAlias, Cancel ct)
+	{
+		logger.LogInformation("Resolving content dates in {Index} via pipeline {Pipeline}", indexAlias, PipelineName);
+
+		await operations.UpdateByQueryAsync(indexAlias, PostData.Empty, PipelineName, ct);
+
+		logger.LogInformation("Content date resolution complete for {Index}", indexAlias);
+	}
+
 	private string GenerateStagingName() =>
 		$"{_lookupAlias}-{DateTime.UtcNow:yyyyMMddHHmmss}-{Guid.NewGuid().ToString("N")[..8]}";
 

src/Elastic.Markdown/Exporters/Elasticsearch/ElasticsearchMarkdownExporter.cs

@@ -48,6 +48,10 @@ public partial class ElasticsearchMarkdownExporter : IMarkdownExporter, IDisposa
 	// Content date tracking - enrich policy + pipeline for content_last_updated
 	private readonly ContentDateEnrichment _contentDateEnrichment;
 
+	// Read aliases resolved during StartAsync, used for post-indexing operations
+	private string _lexicalReadAlias = string.Empty;
+	private string _semanticReadAlias = string.Empty;
+
 	// Per-channel running totals for progress logging
 	private int _primaryIndexed;
 	private int _secondaryIndexed;
@@ -208,6 +212,8 @@ private void ConfigureChannelOptions(string label, IngestChannelOptions<Document
 	public async ValueTask StartAsync(Cancel ctx = default)
 	{
 		var orchestratorContext = await _orchestrator.StartAsync(BootstrapMethod.Failure, ctx);
+		_lexicalReadAlias = orchestratorContext.PrimaryReadAlias;
+		_semanticReadAlias = orchestratorContext.SecondaryReadAlias;
 
 		_logger.LogInformation(
 			"Orchestrator started — strategy: {Strategy}, primary: {PrimaryAlias}, secondary: {SecondaryAlias}",
@@ -218,7 +224,14 @@ public async ValueTask StartAsync(Cancel ctx = default)
 	public async ValueTask StopAsync(Cancel ctx = default)
 	{
 		_ = await _orchestrator.CompleteAsync(null, ctx);
-		await _contentDateEnrichment.SyncLookupIndexAsync(_lexicalTypeContext.IndexStrategy!.WriteTarget!, ctx);
+
+		// Resolve content_last_updated for documents where the ingest pipeline didn't fire.
+		// HashedBulkUpdate uses bulk update actions, which skip ingest pipelines.
+		// Use the read alias (-latest) rather than WriteTarget, which is removed after CompleteAsync.
+		await _contentDateEnrichment.ResolveContentDatesAsync(_lexicalReadAlias, ctx);
+		await _contentDateEnrichment.ResolveContentDatesAsync(_semanticReadAlias, ctx);
+
+		await _contentDateEnrichment.SyncLookupIndexAsync(_lexicalReadAlias, ctx);
 	}
 
 	private async Task PostCompleteAsync(OrchestratorContext<DocumentationDocument> context, Cancel ctx)