Why should you care?

Clinicians need faster handoffs; hospitals need BAAs, audit, and guardrails on every AI touchpoint.

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.Hospital/HospitalAssistantService.cs
public sealed class HospitalAssistantService(
    AllergyPlugin allergies, IChatClient chat, IInputValidator validator)
{
    public async Task<TriageSuggestion> TriageAsync(TriageRequest req, CancellationToken ct)
    {
        validator.ValidateNoPhiInFreeText(req.ChiefComplaint);
        var allergyCtx = await allergies.ListAsync(req.Mrn, ct);
        var prompt = $"Chief complaint: {req.ChiefComplaint}\nAllergies: {allergyCtx}\nSuggest urgency 1-5 with rationale.";
        var text = (await chat.GetResponseAsync([prompt], cancellationToken: ct)).Text;
        return TriageSuggestion.Parse(text);
    }
}

What happened?

• HospitalAssistantService validates input, loads allergies via plugin, returns structured TriageSuggestion — not treatment orders.
• Follow the steps below — typing the code yourself is the fastest way to learn.

Practice next

1. Sign BAA with Azure OpenAI; disable public model training.
2. Run HospitalAssistant behind nurse SSO and audit.
3. HumanReview workflow before EHR writeback.
4. Add urgency enum validation rejecting out-of-range model output.
5. Integrate FHIR read-only client instead of custom MRN repo.

Remember

Healthcare AI assists triage and docs with PHI controls. Plugins supply authoritative allergy and lab context. Human clinicians approve before EHR updates.

ED triage assist

Nurse enters chief complaint; assistant suggests urgency with allergy context.

Outcome: Nurse confirms or overrides; audit stores suggestion and final decision.