Used AI to find bugs and fix them (all changes manually reviewed)

.gitignore

@@ -9,3 +9,5 @@ TestResults/
 
 docs/site
 .cache
+
+coveragereport

CHANGELOG.md

@@ -5,6 +5,31 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## v2.2.1
+
+Add:
+
+- Added `src/test.sh` helper script to run the solution tests with XPlat code coverage and generate HTML, Markdown, and text coverage reports.
+- Added regression tests in `BugDiscoveryTests` and `DiskVectorDatabaseTests` to cover paging metadata, empty database stream round-tripping, disk persistence reload behavior, immediate visibility of disk-backed writes, and immediate reopen-after-delete persistence.
+- Added `CoverageExpansionTests` to extend coverage across model types, exception hierarchy, ID generators, `VectorTextResult`, `TextDataLoader`, `DatabaseFile`, memory store branches, and obsolete serialization wrappers.
+- Added `DiskStoreRegressionTests` to validate disk-backed vocabulary and vector store serialization, deserialization, enumeration, and persistence behavior.
+
+Fixed:
+
+- Fixed `.Search()` / `.SearchAsync()` result paging metadata so `TotalPages` is calculated from the total result count instead of echoing the requested page size.
+- Fixed `MemoryVectorDatabase<TMetadata>` deserialization from an empty binary stream so ID generation resets correctly instead of failing when no items exist.
+- Fixed `BasicDiskVectorStore` read-after-write behavior so added items are immediately visible to `Count`, `GetIds()`, `ContainsKey()`, enumeration, and search before the background disk flush completes.
+- Fixed `BasicDiskVectorStore` delete persistence to avoid WAL truncation races when a database is reopened immediately after delete operations.
+- Fixed reopened disk-backed databases to correctly expose persisted IDs and search results after checkpoint recovery.
+- Fixed `BasicDiskVocabularyStore.DeserializeFromJsonStreamAsync` so deserialized vocabularies restore the in-memory lookup cache as well as the persisted vocabulary map.
+- Fixed `BasicDiskVectorStore.SerializeToJsonStreamAsync` to flush pending operations before serialization and persist actual vector items rather than only index offsets.
+- Fixed `BasicDiskVectorStore.DeserializeFromJsonStreamAsync` to rebuild item storage, index, visible IDs, and cache from serialized vector items.
+- Updated the package version to `2.2.1` and refreshed the copyright year range to `2024-2026`.
+
+Notes:
+
+- Added `coveragereport` to `.gitignore` to keep generated coverage artifacts out of source control.
+
 ## v2.2.0
 
 Add:

src/Build5Nines.SharpVector/Build5Nines.SharpVector.csproj

@@ -9,9 +9,9 @@
     <PackageId>Build5Nines.SharpVector</PackageId>
     <PackageProjectUrl>https://sharpvector.build5nines.com</PackageProjectUrl>
     <RepositoryUrl>https://github.com/Build5Nines/SharpVector</RepositoryUrl>
-    <Version>2.2.0</Version>
+    <Version>2.2.1</Version>
     <Description>Lightweight In-memory Vector Database to embed in any .NET Applications</Description>
-    <Copyright>Copyright (c) 2025 Build5Nines LLC</Copyright>
+    <Copyright>Copyright (c) 2024-2026 Build5Nines LLC</Copyright>
     <PackageReadmeFile>README.md</PackageReadmeFile>
     <PackageLicenseFile>LICENSE</PackageLicenseFile>
     <Authors>Chris Pietschmann</Authors>

src/Build5Nines.SharpVector/MemoryVectorDatabase.cs

@@ -59,8 +59,11 @@ public override async Task DeserializeFromBinaryStreamAsync(Stream stream)
     {
         await base.DeserializeFromBinaryStreamAsync(stream);
 
-        // Re-initialize the IdGenerator with the max Id value from the VectorStore
-        _idGenerator = new IntIdGenerator(VectorStore.GetIds().Max());
+        // Re-initialize the IdGenerator with the max Id value from the VectorStore when items exist.
+        var ids = VectorStore.GetIds().ToArray();
+        _idGenerator = ids.Length > 0
+            ? new IntIdGenerator(ids.Max())
+            : new IntIdGenerator();
     }
 
     /// <summary>
@@ -71,7 +74,10 @@ public override void DeserializeFromBinaryStream(Stream stream)
     {
         base.DeserializeFromBinaryStream(stream);
 
-        // Re-initialize the IdGenerator with the max Id value from the VectorStore
-        _idGenerator = new IntIdGenerator(VectorStore.GetIds().Max());
+        // Re-initialize the IdGenerator with the max Id value from the VectorStore when items exist.
+        var ids = VectorStore.GetIds().ToArray();
+        _idGenerator = ids.Length > 0
+            ? new IntIdGenerator(ids.Max())
+            : new IntIdGenerator();
     }
 }

src/Build5Nines.SharpVector/VectorDatabaseBase.cs

@@ -249,14 +249,20 @@ public async Task<IVectorTextResult<TId, TVocabularyKey, TMetadata>> SearchAsync
         var totalCountFoundInSearch = similarities.Count();
 
         IEnumerable<IVectorTextResultItem<TId, TVocabularyKey, TMetadata>> resultsToReturn;
-        if (pageCount != null && pageCount >= 0 && pageIndex >= 0) {
+        int totalPages;
+        if (pageCount != null && pageCount >= 0 && pageIndex >= 0)
+        {
             resultsToReturn = similarities.Skip(pageIndex * pageCount.Value).Take(pageCount.Value);
-        } else {
+            totalPages = pageCount.Value == 0 ? 0 : (int)Math.Ceiling(totalCountFoundInSearch / (double)pageCount.Value);
+        }
+        else
+        {
             // no paging specified, return all results
             resultsToReturn = similarities;
+            totalPages = totalCountFoundInSearch > 0 ? 1 : 0;
         }
 
-        return new VectorTextResult<TId, TVocabularyKey, TMetadata>(totalCountFoundInSearch, pageIndex, pageCount.HasValue ? pageCount.Value : 1, resultsToReturn);
+        return new VectorTextResult<TId, TVocabularyKey, TMetadata>(totalCountFoundInSearch, pageIndex, totalPages, resultsToReturn);
     }
 
     private async Task<IEnumerable<IVectorTextResultItem<TId, TVocabularyKey, TMetadata>>> CalculateVectorComparisonAsync(TVocabularyKey queryText, float? threshold = null, Func<TMetadata?, Task<bool>>? filter = null)
@@ -673,14 +679,20 @@ public async Task<IVectorTextResult<TId, string, TMetadata>> SearchAsync(string
         var totalCountFoundInSearch = similarities.Count();
 
         IEnumerable<IVectorTextResultItem<TId, string, TMetadata>> resultsToReturn;
-        if (pageCount != null && pageCount >= 0 && pageIndex >= 0) {
+        int totalPages;
+        if (pageCount != null && pageCount >= 0 && pageIndex >= 0)
+        {
             resultsToReturn = similarities.Skip(pageIndex * pageCount.Value).Take(pageCount.Value);
-        } else {
+            totalPages = pageCount.Value == 0 ? 0 : (int)Math.Ceiling(totalCountFoundInSearch / (double)pageCount.Value);
+        }
+        else
+        {
             // no paging specified, return all results
             resultsToReturn = similarities;
+            totalPages = totalCountFoundInSearch > 0 ? 1 : 0;
         }
 
-        return new VectorTextResult<TId, string, TMetadata>(totalCountFoundInSearch, pageIndex, pageCount.HasValue ? pageCount.Value : 1, resultsToReturn);
+        return new VectorTextResult<TId, string, TMetadata>(totalCountFoundInSearch, pageIndex, totalPages, resultsToReturn);
     }
 
     private async Task<IEnumerable<IVectorTextResultItem<TId, string, TMetadata>>> CalculateVectorComparisonAsync(string queryText, float? threshold = null, Func<TMetadata?, Task<bool>>? filter = null)