Why should you care?

Semantic search over ERP help docs and CRM emails requires embeddings before vector storage.

See it live — copy this example

Paste into an ASP.NET Core 8+ / AgentNest project, then run with dotnet run (set your API keys in user-secrets).

// AgentNest.Rag/EmbeddingIndexer.cs
public sealed class EmbeddingIndexer(IEmbeddingGenerator embedder, IVectorStore store)
{
    public async Task IndexAsync(string tenantId, IEnumerable<KnowledgeChunk> chunks, CancellationToken ct)
    {
        foreach (var chunk in chunks)
        {
            var embedding = await embedder.GenerateEmbeddingAsync(chunk.Text, cancellationToken: ct);
            await store.UpsertAsync(tenantId, chunk.Id, embedding.Vector, chunk.Text, ct);
        }
    }
}

What happened?

• EmbeddingIndexer loops chunks, generates vectors via IEmbeddingGenerator, and upserts into IVectorStore with tenant scope.
• Follow the steps below — typing the code yourself is the fastest way to learn.

Practice next

1. Register IEmbeddingGenerator from Azure OpenAI in Program.cs.
2. Define KnowledgeChunk with Id, Text, SourceUri.
3. Run IndexAsync from a nightly hosted service.
4. Batch embed 100 chunks per API call where SDK supports it.
5. Store embedding model name in vector metadata.

Remember

Embeddings power semantic similarity in AgentNest RAG. Index chunked text, not entire files. Track model version when reindexing tenant data.

Hospital protocol index

Clinical team uploads 200 triage protocols for RAG.

Outcome: EmbeddingIndexer builds searchable vectors nightly after uploads.