🔍 Feedback Wanted: VictorLib v1.0.11-FinalBeta – Looking for input before stable release

[IRodriguez13]

Hi everyone 👋

I'm preparing the next version of VictorLib, a C# SDK that wraps a native vector database written in C.

This new release (v1.0.11-FinalBeta) introduces several improvements focused on developer experience, snapshot serialization, and a double-using design pattern for FLAT → HNSW workflows.

🆕 Highlights

• ✅ DumpToPath_snapshot: serializes a vector index to JSON with support for custom paths
• ✅ ReadSnapshot: loads vector data and metadata for reinsertion
• ✅ Double-using pattern: create a FLAT index → dump it → load it as HNSW for fast search
• ✅ Improved error handling with meaningful exceptions
• ✅ Support for both Windows and Linux via internal factory
• ✅ Full XML documentation for API methods
• 🧪 Added 15+ NUnit tests including benchmarks and integration scenarios

---

📦 NuGet

Package: VictorLib
Current: 1.0.10-FinalBeta
Next: 1.0.11-FinalBeta

---

🔧 Repo

📂 GitHub: VictorCsharpBinding-SDK

---

🧪 Example Test (double-using)

[Test]
public void DoubleUsing_DumpFlat_ThenLoadHNSW()
{
    ushort dims = 128;
    string finalPath;

    using (VictorSDK flat = new(IndexType.FLAT, DistanceMethod.COSINE, dims))
    {
        for (ulong i = 1; i <= 100; i++)
        {
            float[] vector = Enumerable.Repeat((float)i / 100, dims).ToArray();
            flat.Insert(i, vector, dims);
        }

        VictorPersistence.SetBasePath(@"D:\Users\pc\Desktop\Indices");
        finalPath = VictorPersistence.DumpToPath_snapshot(flat);
    }

    var snapshot = VictorPersistence.ReadSnapshot(finalPath);

    using (VictorSDK hnsw = new(IndexType.HNSW, snapshot.Method, snapshot.Dimensions, HNSWContext.Create()))
    {
        foreach (var entry in snapshot.Vectors)
            hnsw.Insert(entry.Id, entry.Vector, snapshot.Dimensions);

        float[] query = Enumerable.Repeat(0.33f, dims).ToArray();
        var result = hnsw.Search(query, dims);

        Assert.IsTrue(result.Label > 0);
    }
}

📣 Looking for feedback on:

• API design: does it feel idiomatic in .NET?

• Error handling: is exception usage intuitive?

• Dump/load workflow: clear enough?

• Performance concerns?

• Anything else you'd like to see!

---

• If you’ve tested the SDK or plan to integrate it, I’d love to hear your thoughts — especially before shipping the stable release.

Thanks in advance!