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

package camera

import (
	"fmt"
	"os/exec"
	"path/filepath"
	"regexp"
	"runtime"
	"sort"
	"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)
//   - Linux:   Video4Linux2 (v4l2, enumerated from /dev/video*)
func ListFFmpegDevices() []CameraInfo {
	// Linux 走 /dev/video* 列舉、不依賴 ffmpeg binary（-list_devices 在部分 v4l2 build
	// 不穩定），故不要求 DetectFFmpeg 也能列出裝置節點；實際抓 frame 時才需要 ffmpeg。
	if runtime.GOOS == "linux" {
		return listV4L2Devices()
	}

	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
}

// --- Linux (Video4Linux2) ---

// v4l2DeviceGlob 是列舉 v4l2 攝影機節點的 glob pattern。抽成變數讓測試可覆寫成
// 假的目錄，不依賴實機 /dev。
var v4l2DeviceGlob = "/dev/video*"

// v4l2DeviceIndexRe 從 /dev/video<N> 取出 index N。
var v4l2DeviceIndexRe = regexp.MustCompile(`video(\d+)$`)

// listV4L2Devices 列舉 Linux 上的 v4l2 攝影機節點（/dev/video*）。
//
// 為何用 glob 而非 ffmpeg -list_devices：v4l2 indev 的 -list_devices 支援度依 ffmpeg
// 版本而異、部分 build 不輸出可解析清單；直接列舉 /dev/video* 節點是最穩定的跨版本做法。
//
// 注意：/dev/video* 也包含非攝影機的 V4L2 節點（如 metadata / output device），這裡先
// 全數列出、由使用者選擇；抓 frame 失敗會在 ffmpeg 層以明確錯誤浮現（camera 修 bug 後不再吞錯）。
func listV4L2Devices() []CameraInfo {
	matches, err := filepath.Glob(v4l2DeviceGlob)
	if err != nil {
		return nil
	}
	return parseV4L2Devices(matches)
}

// parseV4L2Devices 把 /dev/video* 路徑清單轉成 CameraInfo（依 index 排序）。
// 抽出來讓測試可直接餵路徑清單、不碰檔案系統。
func parseV4L2Devices(paths []string) []CameraInfo {
	// 依裝置 index 數值排序（Glob 回傳為字典序，video10 會排在 video2 前，需正規化）。
	sorted := append([]string(nil), paths...)
	sort.Slice(sorted, func(i, j int) bool {
		return v4l2Index(sorted[i]) < v4l2Index(sorted[j])
	})

	var cameras []CameraInfo
	for _, path := range sorted {
		idx := v4l2Index(path)
		if idx < 0 {
			continue // 不是 /dev/video<N> 形式、略過
		}
		cameras = append(cameras, CameraInfo{
			ID:     fmt.Sprintf("cam-%d", idx),
			Name:   path, // Linux 以裝置節點路徑作為名稱（buildCaptureArgs 用 index 組回路徑）
			Index:  idx,
			Width:  640,
			Height: 480,
		})
	}
	return cameras
}

// v4l2Index 從 /dev/video<N> 取出 N；非該形式回 -1。
func v4l2Index(path string) int {
	m := v4l2DeviceIndexRe.FindStringSubmatch(path)
	if len(m) != 2 {
		return -1
	}
	n, err := strconv.Atoi(m[1])
	if err != nil {
		return -1
	}
	return n
}