Why should you care?

Developers need fast inner-loop RAG testing without cloud vector bills or VPN to production Pinecone.

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.Dev/ChromaVectorStore.cs — local dev only
public sealed class ChromaVectorStore : IVectorStore
{
    readonly HttpClient _http = new() { BaseAddress = new Uri("http://localhost:8000") };
    public async Task UpsertAsync(string tenantId, string id, ReadOnlyMemory<float> vector, string text, CancellationToken ct)
    {
        var body = new { ids = new[] { id }, embeddings = new[] { vector.ToArray() },
            metadatas = new[] { new { tenant = tenantId } }, documents = new[] { text } };
        await _http.PostAsJsonAsync($"/api/v1/collections/agentnest-{tenantId}/add", body, ct);
    }
}

What happened?

• ChromaVectorStore calls Chroma REST API with per-tenant collection names for local AgentNest CRM experiments.
• Follow the steps below — typing the code yourself is the fastest way to learn.

Practice next

1. pip install chromadb && chroma run --path ./chroma-data.
2. Point VectorStore:Provider=Chroma in appsettings.Development.json.
3. Seed sample CRM chunks via EmbeddingIndexer.
4. Script docker-compose with Chroma + AgentNest.Api.
5. Compare recall@5 against Pinecone on same seed set.

Remember

Chroma accelerates AgentNest local RAG development. Keep collection naming aligned with production adapter. Do not use embedded Chroma for multi-node production.

Developer inner loop

New hire builds CRM RAG feature on laptop without cloud keys.

Outcome: Chroma local index validates chunking before PR merges to shared Pinecone.