visionA/local-agent/server/internal/camera/ffmpeg_detect.go

package camera

import (
	"fmt"
	"os/exec"
	"regexp"
	"runtime"
	"strconv"
	"strings"
)

// DetectFFmpeg checks if ffmpeg is available on the system.
func DetectFFmpeg() bool {
	_, err := exec.LookPath("ffmpeg")
	return err == nil
}

// ListFFmpegDevices detects available video devices using ffmpeg.
// Automatically selects the correct capture framework for the current OS:
//   - macOS: AVFoundation
//   - Windows: DirectShow (dshow)
func ListFFmpegDevices() []CameraInfo {
	if !DetectFFmpeg() {
		return nil
	}

	switch runtime.GOOS {
	case "windows":
		return listDShowDevices()
	default:
		return listAVFoundationDevices()
	}
}

// --- macOS (AVFoundation) ---

func listAVFoundationDevices() []CameraInfo {
	cmd := exec.Command("ffmpeg", "-f", "avfoundation", "-list_devices", "true", "-i", "")
	output, _ := cmd.CombinedOutput()
	return parseAVFoundationOutput(string(output))
}

// parseAVFoundationOutput parses ffmpeg AVFoundation device listing.
// Example:
//
//	[AVFoundation indev @ 0x...] AVFoundation video devices:
//	[AVFoundation indev @ 0x...] [0] FaceTime HD Camera
//	[AVFoundation indev @ 0x...] [1] Capture screen 0
//	[AVFoundation indev @ 0x...] AVFoundation audio devices:
func parseAVFoundationOutput(output string) []CameraInfo {
	var cameras []CameraInfo
	lines := strings.Split(output, "\n")

	deviceRe := regexp.MustCompile(`\[AVFoundation[^\]]*\]\s*\[(\d+)\]\s*(.+)`)

	inVideoSection := false
	for _, line := range lines {
		if strings.Contains(line, "AVFoundation video devices") {
			inVideoSection = true
			continue
		}
		if strings.Contains(line, "AVFoundation audio devices") {
			break
		}
		if !inVideoSection {
			continue
		}

		matches := deviceRe.FindStringSubmatch(line)
		if len(matches) == 3 {
			index, err := strconv.Atoi(matches[1])
			if err != nil {
				continue
			}
			name := strings.TrimSpace(matches[2])

			// Skip screen capture devices
			if strings.Contains(strings.ToLower(name), "capture screen") {
				continue
			}

			cameras = append(cameras, CameraInfo{
				ID:     fmt.Sprintf("cam-%d", index),
				Name:   name,
				Index:  index,
				Width:  640,
				Height: 480,
			})
		}
	}

	return cameras
}

// --- Windows (DirectShow) ---

func listDShowDevices() []CameraInfo {
	cmd := exec.Command("ffmpeg", "-f", "dshow", "-list_devices", "true", "-i", "dummy")
	output, _ := cmd.CombinedOutput()
	return parseDShowOutput(string(output))
}

// parseDShowOutput parses ffmpeg DirectShow device listing.
// Example:
//
//	[dshow @ 0x...] "Integrated Camera" (video)
//	[dshow @ 0x...]   Alternative name "@device_pnp_..."
//	[dshow @ 0x...] "Microphone" (audio)
func parseDShowOutput(output string) []CameraInfo {
	var cameras []CameraInfo
	lines := strings.Split(output, "\n")

	// Match: [dshow @ 0x...] "Device Name" (video)
	deviceRe := regexp.MustCompile(`\[dshow[^\]]*\]\s*"([^"]+)"\s*\(video\)`)

	index := 0
	for _, line := range lines {
		matches := deviceRe.FindStringSubmatch(line)
		if len(matches) == 2 {
			name := strings.TrimSpace(matches[1])

			cameras = append(cameras, CameraInfo{
				ID:     fmt.Sprintf("cam-%d", index),
				Name:   name,
				Index:  index,
				Width:  640,
				Height: 480,
			})
			index++
		}
	}

	return cameras
}