Why should you care?

Hospital staff query protocols in natural language; keyword SQL LIKE fails on synonyms and abbreviations.

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/SemanticSearchService.cs
public sealed class SemanticSearchService(IEmbeddingGenerator embedder, IVectorStore store)
{
    public async Task<IReadOnlyList<SearchHit>> QueryAsync(
        string tenantId, string query, int topK, CancellationToken ct)
    {
        var q = await embedder.GenerateEmbeddingAsync(query, cancellationToken: ct);
        var hits = await store.SearchAsync(tenantId, q.Vector, topK, ct);
        return hits.Select(h => new SearchHit(h.Id, h.Score, h.Text, h.Source)).ToList();
    }
}

What happened?

• SemanticSearchService embeds the query vector and returns ranked SearchHit DTOs with score and source for UI highlighting.
• Follow the steps below — typing the code yourself is the fastest way to learn.

Practice next

1. Expose GET /api/search/semantic?q=... with tenant from JWT.
2. Set minimum score threshold to filter weak matches.
3. Combine with BM25 hybrid search optional for SKU codes.
4. Add hybrid search weight slider in admin UI.
5. Log zero-hit queries to improve chunk coverage.

Remember

Semantic search matches by embedding similarity. Expose scores and sources in SearchHit DTOs. Tune topK and thresholds per AgentNest module.

CRM deal discovery

Rep searches renewal risk and finds deals mentioning churn signals indirectly.

Outcome: Semantic search surfaces relevant opportunities keyword search missed.