Runtime/AgentCore/Tools/Builtin/CountVisibleObjectsTool.cs

using System;
using System.Collections.Generic;
using System.Text;
using System.Threading.Tasks;
using Microsoft.Extensions.AI;
using OnDeviceAgent.Inference;
namespace OnDeviceAgent.AgentCore
{

    public sealed class CountVisibleObjectsTool : AgentToolBase
    {
        readonly CameraInputService m_Camera;
        readonly YoloDetector m_Yolo;

        public CountVisibleObjectsTool(CameraInputService camera, YoloDetector yolo)
        {
            m_Camera = camera;
            m_Yolo = yolo;
        }

        public override string Name => "CountVisibleObjects";
        public override string Description =>
            "Count how many of something are in the live camera view right now (e.g. 'how many people do you see', '사람 몇 명 보여').";

        public override AITool ToAIFunction(AgentToolContext context)
        {
            Func<Task<string>> body = null;
            if (m_Camera != null && m_Yolo != null)
                body = async () =>
                {
                    // WebCamTexture + Worker.Schedule are main-thread only.
                    var detections = await context.Dispatcher.RunOnMainAsync<List<YoloDetector.Detection>>(async () =>
                    {
                        if (!m_Camera.HasFrame) return null;
                        // Surface the same frame in the result window (fires FrameJpegCaptured), matching the
                        // vision tool — so the user sees what was counted, not just the number.
                        m_Camera.GetLatestFrameJpeg();
                        if (!m_Yolo.IsReady) return new List<YoloDetector.Detection>();
                        return await m_Yolo.DetectAsync(m_Camera.GetLatestFrameTexture()).ConfigureAwait(false);
                    }).ConfigureAwait(false);
                    return detections == null ? "The camera has no frame yet." : Summarize(detections);
                };

            return AIFunctionFactory.Create(
                new Func<Task<string>>(() => RunGuarded(context, "{}", body, "Vision is not configured.")),
                name: Name,
                description: Description);
        }

        static string Summarize(List<YoloDetector.Detection> detections)
        {
            if (detections == null || detections.Count == 0)
                return "No recognizable objects are visible in the current camera view.";

            var counts = new Dictionary<string, int>(StringComparer.Ordinal);
            foreach (var d in detections)
            {
                counts.TryGetValue(d.ClassName, out var n);
                counts[d.ClassName] = n + 1;
            }

            var ordered = new List<KeyValuePair<string, int>>(counts);
            ordered.Sort((a, b) =>
            {
                if (a.Key == "person" && b.Key != "person") return -1;
                if (b.Key == "person" && a.Key != "person") return 1;
                return b.Value.CompareTo(a.Value);
            });

            var sb = new StringBuilder("Currently visible: ");
            for (var i = 0; i < ordered.Count; i++)
            {
                if (i > 0) sb.Append(", ");
                sb.Append(ordered[i].Value).Append(' ').Append(Pluralize(ordered[i].Key, ordered[i].Value));
            }
            sb.Append('.');
            return sb.ToString();
        }

        static string Pluralize(string noun, int count)
        {
            if (count <= 1) return noun;
            if (noun == "person") return "people";
            if (noun.EndsWith("s") || noun.EndsWith("x") || noun.EndsWith("ch") || noun.EndsWith("sh"))
                return noun + "es";
            return noun + "s";
        }
    }
}