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

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

app/MindWork AI Studio/Tools/Services/DataSourceEmbeddingManifest.cs

@@ -1,7 +1,27 @@
 namespace AIStudio.Tools.Services;
 
+/// <summary>
+/// What the index store holds for one data source, read at the start of a run.
+/// </summary>
 public sealed class DataSourceEmbeddingManifest
 {
+    /// <summary>
+    /// How the index compares the paths of the documents it stores, on every OS.
+    /// </summary>
+    /// <remarks>
+    /// Ignores case, unlike PathTools.COMPARER, which follows the file system. The index store keeps
+    /// the paths with the NOCASE collation and allows each path of a data source once, so two files
+    /// which differ in their case alone share one row there. Whatever the indexer keys or compares by
+    /// these paths has to agree with the store; telling the two files apart would let them replace
+    /// each other's row on every run.
+    /// </remarks>
+    public static readonly StringComparer PATH_COMPARER = StringComparer.OrdinalIgnoreCase;
+
+    /// <summary>
+    /// The comparison matching PATH_COMPARER, e.g., for prefixes.
+    /// </summary>
+    public const StringComparison PATH_COMPARISON = StringComparison.OrdinalIgnoreCase;
+
     public string EmbeddingProviderId { get; set; } = string.Empty;
 
     public string EmbeddingSignature { get; set; } = string.Empty;
@@ -10,7 +30,7 @@ public sealed class DataSourceEmbeddingManifest
 
     public int VectorSize { get; set; }
 
-    public Dictionary<string, EmbeddedFileRecord> Files { get; init; } = new(StringComparer.OrdinalIgnoreCase);
+    public Dictionary<string, EmbeddedFileRecord> Files { get; init; } = new(PATH_COMPARER);
 
     /// <summary>
     /// The files whose indexing began but never came to an end, by their absolute path.
@@ -21,7 +41,7 @@ public sealed class DataSourceEmbeddingManifest
     /// meanwhile. Indexing such a file again replaces both. A file which is gone, or which the data
     /// source no longer selects, is never indexed again, so the indexer has to remove them itself.
     /// </remarks>
-    public HashSet<string> UnfinishedFiles { get; init; } = new(StringComparer.OrdinalIgnoreCase);
+    public HashSet<string> UnfinishedFiles { get; init; } = new(PATH_COMPARER);
 
     /// <summary>
     /// The files whose indexing failed for a reason which lies in the file itself, keyed by their
@@ -32,5 +52,5 @@ public sealed class DataSourceEmbeddingManifest
     /// a folder holding hundreds of scanned documents without a text layer would be read again on
     /// every single run, with the outcome known in advance.
     /// </remarks>
-    public Dictionary<string, PermanentIndexingFailureRecord> PermanentFailures { get; init; } = new(StringComparer.OrdinalIgnoreCase);
+    public Dictionary<string, PermanentIndexingFailureRecord> PermanentFailures { get; init; } = new(PATH_COMPARER);
 }