Allowed moving local data sources without embedding them again (#1091)

Co-authored-by: Thorsten Sommer <SommerEngineering@users.noreply.github.com>

app/Tests/Tools/Databases/IndexStoreRelocationTests.cs

@@ -0,0 +1,47 @@
+using AIStudio.Tools.Databases.IndexStore;
+using AIStudio.Tools.Services.Indexing;
+
+namespace AIStudio.Tests.Tools.Databases;
+
+public sealed class IndexStoreRelocationTests
+{
+    [Test]
+    public async Task RelocationPreservesChunksAndTranscriptAndUpdatesSearchLocations()
+    {
+        await using var store = await TemporaryIndexStore.CreateAsync();
+        var sourceId = Guid.NewGuid().ToString();
+        var oldPath = Path.Combine(Path.GetTempPath(), "original", "notes.md");
+        var newPath = Path.Combine(Path.GetTempPath(), "moved", "renamed.md");
+        var parentId = IndexedDocumentIds.CreateParentId(sourceId, oldPath);
+        var chunkId = IndexedDocumentIds.CreateChunkId(sourceId, "old-fingerprint", 0);
+        var now = DateTimeOffset.UtcNow;
+        var token = CancellationToken.None;
+        await store.Client.UpsertDataSourceAsync(sourceId, "LOCAL_DIRECTORY", "provider", "signature", "old-hash", 3, token);
+        await store.Client.UpsertFileAsync(sourceId, new EmbeddingStateFile(parentId, oldPath, "notes.md", "notes.md", "md", "old-fingerprint", 100, now, now, now, 1) { TranscriptKey = "transcript" }, token);
+        await store.Client.UpsertChunksAsync(sourceId, [new EmbeddingStateChunk(chunkId, parentId, 2, 0, "report about relocation", now)], token);
+
+        // Repeating the metadata write after an interrupted relocation must keep the same chunks.
+        await store.Client.RelocateFileAsync(sourceId, parentId, newPath, "renamed.md", "new-fingerprint", token);
+        await store.Client.RelocateFileAsync(sourceId, parentId, newPath, "renamed.md", "new-fingerprint", token);
+        var manifest = await store.Client.GetManifestAsync(sourceId, token);
+        var hits = await store.Client.SearchChunksAsync(sourceId, "relocation", 10, token);
+
+        Assert.Multiple(() =>
+        {
+            Assert.That(manifest.Files.Keys, Is.EquivalentTo(new[] { newPath }));
+            Assert.That(manifest.Files[newPath].Fingerprint, Is.EqualTo("new-fingerprint"));
+            Assert.That(manifest.Files[newPath].ParentFileId, Is.EqualTo(parentId));
+            Assert.That(manifest.Files[newPath].RelativePath, Is.EqualTo("renamed.md"));
+            Assert.That(manifest.Files[newPath].TranscriptKey, Is.EqualTo("transcript"));
+            Assert.That(hits, Has.Count.EqualTo(1));
+            Assert.That(hits[0].ChunkId, Is.EqualTo(chunkId));
+            Assert.That(hits[0].AbsolutePath, Is.EqualTo(newPath));
+            Assert.That(hits[0].FileName, Is.EqualTo("renamed.md"));
+            Assert.That(hits[0].PageNumber, Is.EqualTo(2));
+            Assert.That(hits[0].ChunkText, Is.EqualTo("report about relocation"));
+        });
+
+        await store.Client.DeleteFileAsync(sourceId, newPath, token);
+        Assert.That(await store.Client.SearchChunksAsync(sourceId, "relocation", 10, token), Is.Empty, "Ordinary removal must still find a relocated document's chunks.");
+    }
+}