diff --git a/AD_FILTER_GUIDE.md b/AD_FILTER_GUIDE.md new file mode 100644 index 0000000..cd264a5 --- /dev/null +++ b/AD_FILTER_GUIDE.md @@ -0,0 +1,118 @@ +# 🛡️ 广告过滤功能使用指南 + +KVideo 新增了强大的 M3U8 广告过滤系统，支持多种过滤模式，可有效去除视频中插入的切片广告。 + +## ✨ 功能特性 + +- **多模式选择**：支持关闭、关键词过滤、智能启发式过滤(Beta)和激进模式。 +- **UI 集成**：在播放器设置菜单中直接切换模式，实时生效。 +- **自定义关键词**：支持通过环境变量扩展过滤关键词。 +- **高性能**：基于流式处理，对播放加载速度几乎无影响。 + +## 🎮 使用说明 + +1. 打开任意视频播放。 + +2. 点击播放器左上角的 **(···)** 按钮。 + +3. 在 **广告过滤** 下拉菜单中选择模式： + - **关闭**：不过滤任何内容（默认）。 + - **关键词**：仅过滤 URL 中包含特定广告关键词（如 `adjump`, `pre-roll`）的片段。安全稳定，误杀率极低。 + - **智能(Beta)**：推荐使用。结合关键词、文件名特征、时间戳跳变标记 (DISCONTINUITY) 和 HLS 标签进行综合评分。能识别大多数无明显关键词的广告。 + - **激进**：降低判定阈值，过滤更多疑似广告，但可能会误伤正片。仅在智能模式无效时尝试。 + + > ⚠️重要提示：添加关键词时，请勿使用类似`ad` `ads` `video` `20260118` 这类简短相对通用的词汇，会造成误杀率极高。 + +--- + +## 🛠️ 部署配置指南 + +你可以通过环境变量 `NEXT_PUBLIC_AD_KEYWORDS` 自定义额外的广告关键词。 + +### 1. 本地开发 (Local Development) + +在项目根目录的 `.env.local` 文件中添加。支持多行格式（推荐使用引号包裹）： + +```env +# 单行模式 +NEXT_PUBLIC_AD_KEYWORDS=spam,promo,intro-ad + +# 多行模式 (更清晰) +NEXT_PUBLIC_AD_KEYWORDS=" +spam +promo +intro-ad +/unwanted-segment/ +" +``` + +### 2. Docker 部署 (运行时热加载) + +我们支持三种方式注入关键词，优先级从高到低： + +1. **文件加载 (推荐)**：通过挂载文件并在运行时读取。 +2. **运行时环境变量**：直接通过 `-e` 传入 `AD_KEYWORDS`。 +3. **构建时环境变量**：构建镜像时打入的 `NEXT_PUBLIC_AD_KEYWORDS`。 + +#### 方案 A：挂载配置文件 (支持运行时修改) + +创建一个关键词文件 `ad_keywords.txt`： +```text +spam +promo +/ad-segment/ +intro +``` + +运行 Docker 时挂载该文件并设置 `AD_KEYWORDS_FILE` 环境变量： + +```bash +docker run -d -p 3000:3000 \ + -v $(pwd)/ad_keywords.txt:/app/ad_keywords.txt \ + -e AD_KEYWORDS_FILE=/app/ad_keywords.txt \ + --name kvideo kuekhaoyang/kvideo:latest +``` +*注：修改宿主机文件后重启容器即可生效，无需重新构建镜像。* + +#### 方案 B：直接注入环境变量 + +```bash +docker run -d -p 3000:3000 \ + -e AD_KEYWORDS="spam,promo,intro-ad" \ + --name kvideo kuekhaoyang/kvideo:latest +``` + +### 3. Vercel 部署 + +1. 进入 Vercel 项目 Dashboard。 +2. 点击 **Settings** -> **Environment Variables**。 +3. 添加新变量： + - **Key**: `NEXT_PUBLIC_AD_KEYWORDS` + - **Value**: `spam,promo,intro-ad` +4. 重新部署项目以生效。 + +### 4. Cloudflare Pages 部署 + +1. 进入 Cloudflare Pages 项目 Dashboard。 +2. 点击 **Settings** -> **Environment variables**。 +3. 添加变量： + - **Variable name**: `NEXT_PUBLIC_AD_KEYWORDS` + - **Value**: `spam,promo,intro-ad` +4. 重新部署项目 (Retry deployment) 以生效。 + +--- + +## 🔍 技术实现细节 + +广告检测逻辑位于 `lib/utils/m3u8-ad-detector.ts`，采用分块评分机制： + +1. **Block 解析**：将 M3U8 播放列表按 `#EXT-X-DISCONTINUITY` 分割成多个块。 +2. **特征学习**：分析最长的块（通常是正片）提取特征，如文件命名模式、路径前缀等。 +3. **评分系统**： + - **HLS 标签**：`#EXT-X-CUE-OUT`/`IN` (+10分，确认为广告) + - **路径前缀**：与正片路径不一致 (+5分，高度疑似) + - **关键词**：URL 包含已知广告词 (+2.5分) + - **文件名模式**：与正片命名规则不符 (+1.5分) +4. **决策过滤**： + - **智能模式**：移除评分 >= 5.0 的块。 + - **激进模式**：移除评分 >= 3.0 的块。 diff --git a/README.md b/README.md index 9293b54..e8b0f33 100644 --- a/README.md +++ b/README.md @@ -83,6 +83,11 @@ - **内容隔离**：高级内容与普通内容完全物理隔离，互不干扰 - **专属设置**：拥有独立的内容源管理和功能设置 +### 🛡️ 广告过滤 +- **多模式选择**：支持关闭、关键词过滤、智能启发式过滤(Beta)和激进模式。 +- **UI 集成**：在播放器设置菜单中直接切换模式，实时生效。 +- **自定义关键词**：支持通过环境变量扩展过滤关键词。 +- **高性能**：基于流式处理，对播放加载速度几乎无影响。 ## 🔐 隐私保护 diff --git a/app/layout.tsx b/app/layout.tsx index 73e9ca1..4a89cb3 100644 --- a/app/layout.tsx +++ b/app/layout.tsx @@ -1,3 +1,4 @@ +import React from 'react'; import type { Metadata } from "next"; import { Geist, Geist_Mono } from "next/font/google"; import "./globals.css"; @@ -6,7 +7,48 @@ import { Analytics } from "@vercel/analytics/react"; import { ServiceWorkerRegister } from "@/components/ServiceWorkerRegister"; import { PasswordGate } from "@/components/PasswordGate"; import { siteConfig } from "@/lib/config/site-config"; +import { AdKeywordsInjector } from "@/components/AdKeywordsInjector"; +import fs from 'fs'; +import path from 'path'; +// Server Component specifically for reading env/file (async for best practices) +async function AdKeywordsWrapper() { + let keywords: string[] = []; + + try { + // 1. Try reading from file (Docker runtime support) + const keywordsFile = process.env.AD_KEYWORDS_FILE; + if (keywordsFile) { + // Resolve absolute path or relative to CWD + const filePath = path.isAbsolute(keywordsFile) + ? keywordsFile + : path.join(process.cwd(), keywordsFile); + + try { + const content = await fs.promises.readFile(filePath, 'utf-8'); + keywords = content.split(/[\n,]/).map((k: string) => k.trim()).filter((k: string) => k); + console.log(`[AdFilter] Loaded ${keywords.length} keywords from file: ${filePath}`); + } catch (fileError: unknown) { + // Handle file not found (ENOENT) gracefully + if ((fileError as NodeJS.ErrnoException).code !== 'ENOENT') { + console.warn('[AdFilter] Error reading keywords file:', fileError); + } + } + } + + // 2. Fallback to Env var (Runtime or Build time) + if (keywords.length === 0) { + const envKeywords = process.env.AD_KEYWORDS || process.env.NEXT_PUBLIC_AD_KEYWORDS; + if (envKeywords) { + keywords = envKeywords.split(/[\n,]/).map((k: string) => k.trim()).filter((k: string) => k); + } + } + } catch (error) { + console.warn('[AdFilter] Failed to load keywords:', error); + } + + return ; +} const geistSans = Geist({ variable: "--font-geist-sans", @@ -39,6 +81,7 @@ export default function RootLayout({ > + {children} diff --git a/components/AdKeywordsInjector.tsx b/components/AdKeywordsInjector.tsx new file mode 100644 index 0000000..6c2e161 --- /dev/null +++ b/components/AdKeywordsInjector.tsx @@ -0,0 +1,22 @@ +'use client'; + +import { useEffect, useRef } from 'react'; +import { usePlayerSettings } from '@/components/player/hooks/usePlayerSettings'; + +interface AdKeywordsInjectorProps { + keywords: string[]; +} + +export function AdKeywordsInjector({ keywords }: AdKeywordsInjectorProps) { + const { setAdKeywords } = usePlayerSettings(); + const initialized = useRef(false); + + useEffect(() => { + if (!initialized.current && keywords.length > 0) { + setAdKeywords(keywords); + initialized.current = true; + } + }, [keywords, setAdKeywords]); + + return null; +} diff --git a/components/player/desktop/DesktopMoreMenu.tsx b/components/player/desktop/DesktopMoreMenu.tsx index f736b5c..f7ac66b 100644 --- a/components/player/desktop/DesktopMoreMenu.tsx +++ b/components/player/desktop/DesktopMoreMenu.tsx @@ -3,6 +3,7 @@ import React from 'react'; import { Icons } from '@/components/ui/Icon'; import { usePlayerSettings } from '../hooks/usePlayerSettings'; +import { settingsStore, AdFilterMode } from '@/lib/store/settings-store'; import { createPortal } from 'react-dom'; @@ -32,16 +33,28 @@ export function DesktopMoreMenu({ autoSkipOutro, skipOutroSeconds, showModeIndicator, + adFilter, setAutoNextEpisode, setAutoSkipIntro, setSkipIntroSeconds, setAutoSkipOutro, setSkipOutroSeconds, setShowModeIndicator, + setAdFilter, + adFilterMode, + setAdFilterMode, } = usePlayerSettings(); const buttonRef = React.useRef(null); const [menuPosition, setMenuPosition] = React.useState({ top: 0, left: 0 }); + const [isAdFilterOpen, setAdFilterOpen] = React.useState(false); + + const AD_FILTER_LABELS: Record = { + off: '关闭', + keyword: '关键词', + heuristic: '智能(Beta)', + aggressive: '激进' + }; React.useEffect(() => { if (showMoreMenu && buttonRef.current && containerRef.current) { @@ -130,6 +143,46 @@ export function DesktopMoreMenu({ + {/* Ad Filter Mode Selector */} +
+
+ + 广告过滤 +
+ {/* Custom Ad Filter Mode Selector */} +
+ + + {isAdFilterOpen && ( + <> +
setAdFilterOpen(false)} /> +
+ {Object.entries(AD_FILTER_LABELS).map(([mode, label]) => ( + + ))} +
+ + )} +
+
+ {/* Auto Next Episode Switch */}
@@ -225,7 +278,7 @@ export function DesktopMoreMenu({
)}
-
+ ); return ( diff --git a/components/player/hooks/useHlsPlayer.ts b/components/player/hooks/useHlsPlayer.ts index 136b019..b70f27c 100644 --- a/components/player/hooks/useHlsPlayer.ts +++ b/components/player/hooks/useHlsPlayer.ts @@ -1,5 +1,7 @@ import { useEffect, useRef } from 'react'; import Hls from 'hls.js'; +import { usePlayerSettings } from './usePlayerSettings'; +import { filterM3u8Ad } from '@/lib/utils/m3u8-utils'; interface UseHlsPlayerProps { videoRef: React.RefObject; @@ -17,6 +19,8 @@ export function useHlsPlayer({ onError }: UseHlsPlayerProps) { const hlsRef = useRef(null); + const { adFilterMode, adKeywords } = usePlayerSettings(); + const isAdFilterEnabled = adFilterMode !== 'off'; useEffect(() => { const video = videoRef.current; @@ -29,45 +33,66 @@ export function useHlsPlayer({ } let hls: Hls | null = null; + let extraBlobs: string[] = []; // Check if HLS is supported natively (Safari, Mobile Chrome) - // We prefer native playback if available as it's usually more battery efficient const isNativeHlsSupported = video.canPlayType('application/vnd.apple.mpegurl'); if (Hls.isSupported()) { - // Use hls.js for browsers without native support (Desktop Chrome, Firefox, Edge) - // OR if we want to force hls.js for better control (optional, but sticking to native first is safer) - // Note: Some desktop browsers (like Safari) support native HLS. - // We usually prefer native, BUT sometimes native implementation is buggy or lacks features. - // For now, we follow the standard pattern: Native first, then HLS.js. - // EXCEPT for Chrome on Desktop which reports canPlayType as '' (false). + // Define custom loader class to intercept manifest loading + // We use 'any' cast because default loader type might not be strictly exposed in all typings + const DefaultLoader = (Hls as any).DefaultConfig.loader; - if (!isNativeHlsSupported) { - hls = new Hls({ + class AdFilterLoader extends DefaultLoader { + load(context: any, config: any, callbacks: any) { + if (isAdFilterEnabled && (context.type === 'manifest' || context.type === 'level')) { + const originalOnSuccess = callbacks.onSuccess; + callbacks.onSuccess = (response: any, stats: any, context: any, networkDetails: any) => { + if (typeof response.data === 'string') { + try { + // Filter the content + response.data = filterM3u8Ad(response.data, context.url, adFilterMode, adKeywords); + } catch (e) { + console.warn('[HLS] Ad filter error:', e); + } + } + originalOnSuccess(response, stats, context, networkDetails); + }; + } + super.load(context, config, callbacks); + } + } + + if (!isNativeHlsSupported || isAdFilterEnabled) { + // If ad filtering is on, we force Hls.js even on native-supported desktop browsers + // Exceptions might exist for iOS where MSE is strictly not available, check Hls.isSupported() result carefully. + // Hls.isSupported() is false on iOS Safari usually, so this block won't run there. + + const config: any = { // Worker & Performance enableWorker: true, - lowLatencyMode: false, // Disable low latency for more stable playback + lowLatencyMode: false, - // Buffer Settings - More aggressive buffering for smoother playback - maxBufferLength: 60, // Buffer up to 60 seconds ahead - maxMaxBufferLength: 120, // Allow up to 2 minutes of buffer - maxBufferSize: 60 * 1000 * 1000, // 60MB buffer size - maxBufferHole: 0.5, // Allow small gaps in buffer + // Buffer Settings + maxBufferLength: 60, + maxMaxBufferLength: 120, + maxBufferSize: 60 * 1000 * 1000, + maxBufferHole: 0.5, - // Start with more buffer before playing - startFragPrefetch: true, // Enable prefetching next fragment + // Start with more buffer + startFragPrefetch: true, - // ABR (Adaptive Bitrate) Settings - Be more conservative - abrEwmaDefaultEstimate: 500000, // Start with conservative bandwidth estimate (500kbps) - abrEwmaFastLive: 3, // Fast adaptation for live - abrEwmaSlowLive: 9, // Slow adaptation for live - abrEwmaFastVoD: 3, // Fast adaptation for VoD - abrEwmaSlowVoD: 9, // Slow adaptation for VoD - abrBandWidthFactor: 0.8, // Use 80% of estimated bandwidth (conservative) - abrBandWidthUpFactor: 0.7, // Even more conservative when switching up + // ABR Settings + abrEwmaDefaultEstimate: 500000, + abrEwmaFastLive: 3, + abrEwmaSlowLive: 9, + abrEwmaFastVoD: 3, + abrEwmaSlowVoD: 9, + abrBandWidthFactor: 0.8, + abrBandWidthUpFactor: 0.7, - // Loading Settings - More retries and longer timeouts + // Loading Settings fragLoadingMaxRetry: 6, fragLoadingRetryDelay: 1000, fragLoadingMaxRetryTimeout: 64000, @@ -79,29 +104,35 @@ export function useHlsPlayer({ levelLoadingMaxRetryTimeout: 64000, // Timeouts - fragLoadingTimeOut: 20000, // 20 seconds for fragment loading - manifestLoadingTimeOut: 10000, // 10 seconds for manifest - levelLoadingTimeOut: 10000, // 10 seconds for level + fragLoadingTimeOut: 20000, + manifestLoadingTimeOut: 10000, + levelLoadingTimeOut: 10000, - // Backbuffer - Keep some played content for seeking back - backBufferLength: 30, // Keep 30 seconds of played content - }); + // Backbuffer + backBufferLength: 30, + }; + + // Use custom loader if ad filtering is enabled + if (isAdFilterEnabled) { + config.loader = AdFilterLoader; + } + + hls = new Hls(config); hlsRef.current = hls; hls.loadSource(src); hls.attachMedia(video); + // Auto Play Handler hls.on(Hls.Events.FRAG_LOADED, (event, data) => { - // Force play if we have the first segment and it's not playing yet - // detailed: data.frag.sn is the sequence number if (autoPlay && video.paused && data.frag.start === 0) { video.play().catch(console.warn); } }); + // Manifest Parsed Handler hls.on(Hls.Events.MANIFEST_PARSED, () => { - - // Check for HEVC/H.265 codec (limited browser support) + // Check for HEVC if (hls) { const levels = hls.levels; if (levels && levels.length > 0) { @@ -110,10 +141,7 @@ export function useHlsPlayer({ level.videoCodec?.toLowerCase().includes('h265') ); if (hasHEVC) { - console.warn('[HLS] ⚠️ HEVC/H.265 codec detected - may not play in all browsers'); - console.warn('[HLS] Supported: Safari with hardware acceleration, some Edge versions'); - console.warn('[HLS] Not supported: Most Chrome/Firefox versions'); - // Notify parent about potential codec issues + console.warn('[HLS] ⚠️ HEVC detected'); onError?.('检测到 HEVC/H.265 编码，当前浏览器可能不支持'); } } @@ -121,12 +149,13 @@ export function useHlsPlayer({ if (autoPlay) { video.play().catch((err) => { - console.warn('[HLS] Autoplay prevented:', err); + // console.warn('[HLS] Autoplay prevented:', err); onAutoPlayPrevented?.(err); }); } }); + // Error Handling let networkErrorRetries = 0; let mediaErrorRetries = 0; const MAX_RETRIES = 3; @@ -136,22 +165,18 @@ export function useHlsPlayer({ switch (data.type) { case Hls.ErrorTypes.NETWORK_ERROR: networkErrorRetries++; - console.error(`[HLS] Network error (${networkErrorRetries}/${MAX_RETRIES}), trying to recover...`, data); if (networkErrorRetries <= MAX_RETRIES) { hls?.startLoad(); } else { - console.error('[HLS] Too many network errors, giving up'); onError?.('网络错误：无法加载视频流'); hls?.destroy(); } break; case Hls.ErrorTypes.MEDIA_ERROR: mediaErrorRetries++; - console.error(`[HLS] Media error (${mediaErrorRetries}/${MAX_RETRIES}), trying to recover...`, data); if (mediaErrorRetries <= MAX_RETRIES) { hls?.recoverMediaError(); } else { - console.error('[HLS] Too many media errors, giving up'); onError?.('媒体错误：视频格式不支持或已损坏'); hls?.destroy(); } @@ -162,20 +187,142 @@ export function useHlsPlayer({ hls?.destroy(); break; } - } else { - // Non-fatal errors - console.warn('[HLS] Non-fatal error:', data.type, data.details); } }); } else { - // Native HLS support + // Native HLS (Desktop Safari, no Filter) video.src = src; } } else if (isNativeHlsSupported) { - // Fallback for environments where Hls.js is not supported but native is (e.g. iOS without MSE?) - video.src = src; + // Native HLS (iOS, Mobile Safari) + // Limitations: Native HLS cannot easily intercept sub-playlist requests. + // We use fetch+blob for the master playlist as a best 'first-level' filter. + // If the ad discontinuity is in the master playlist (rare for ads, common for periods), it works. + // If it's in sub-playlists, it might fail unless we parse and blob those too (complex). + + if (isAdFilterEnabled) { + const processMasterPlaylist = async (masterSrc: string) => { + // Move blob tracking outside try to ensure cleanup on error + const createdBlobs: string[] = []; + + // Safely resolve relative URLs to absolute (handles iOS Safari scenarios) + let absoluteMasterSrc: string; + try { + absoluteMasterSrc = new URL(masterSrc, window.location.href).toString(); + } catch { + absoluteMasterSrc = masterSrc; // Fallback if URL parsing fails + } + + try { + const response = await fetch(absoluteMasterSrc); + const masterContent = await response.text(); + + // If it's a simple playlist (no variants), just filter and play + if (!masterContent.includes('#EXT-X-STREAM-INF')) { + const filtered = filterM3u8Ad(masterContent, absoluteMasterSrc, adFilterMode, adKeywords); + const blob = new Blob([filtered], { type: 'application/vnd.apple.mpegurl' }); + const blobUrl = URL.createObjectURL(blob); + createdBlobs.push(blobUrl); + return { masterBlobUrl: blobUrl, allBlobs: createdBlobs }; + } + + // It IS a master playlist. Use map + Promise.all for clean concurrent processing. + const lines = masterContent.split(/\r?\n/); + + + + // Process each line, looking back at previous line to determine context + const lineProcessingPromises = lines.map(async (line, index) => { + const trimmedLine = line.trim(); + + // Handle #EXT-X-MEDIA:URI="..." + if (trimmedLine.startsWith('#EXT-X-MEDIA') && trimmedLine.includes('URI="')) { + const uriMatch = trimmedLine.match(/URI="([^"]+)"/); + const uri = uriMatch?.[1]; + if (uri) { + // Process if relative or absolute URL; fetch will handle CORS + const isRelative = !uri.startsWith('http'); + + if (isRelative || uri.startsWith('http')) { + try { + const absoluteUrl = isRelative ? new URL(uri, absoluteMasterSrc).toString() : uri; + const subRes = await fetch(absoluteUrl); + const subContent = await subRes.text(); + const filteredSub = filterM3u8Ad(subContent, absoluteUrl, adFilterMode, adKeywords); + const subBlob = new Blob([filteredSub], { type: 'application/vnd.apple.mpegurl' }); + const subBlobUrl = URL.createObjectURL(subBlob); + createdBlobs.push(subBlobUrl); + return line.replace(`URI="${uri}"`, `URI="${subBlobUrl}"`); + } catch (e) { + console.warn('[HLS Native] Failed to process EXT-X-MEDIA URI:', e); + return line; + } + } + } + } + + // Handle playlist URL (line after #EXT-X-STREAM-INF) + const prevLine = index > 0 ? lines[index - 1].trim() : ''; + if (prevLine.startsWith('#EXT-X-STREAM-INF') && trimmedLine && !trimmedLine.startsWith('#')) { + // Process if relative or absolute URL; fetch will handle CORS + const isRelative = !trimmedLine.startsWith('http'); + + if (isRelative || trimmedLine.startsWith('http')) { + try { + const absoluteUrl = isRelative ? new URL(trimmedLine, absoluteMasterSrc).toString() : trimmedLine; + const subRes = await fetch(absoluteUrl); + const subContent = await subRes.text(); + const filteredSub = filterM3u8Ad(subContent, absoluteUrl, adFilterMode, adKeywords); + const subBlob = new Blob([filteredSub], { type: 'application/vnd.apple.mpegurl' }); + const subBlobUrl = URL.createObjectURL(subBlob); + createdBlobs.push(subBlobUrl); + return subBlobUrl; + } catch (e) { + console.warn('[HLS Native] Failed to process variant playlist:', e); + return line; + } + } + } + + // All other lines pass through unchanged + return line; + }); + + const processedLines = await Promise.all(lineProcessingPromises); + + // Join back + const finalMasterContent = processedLines.join('\n'); + const masterBlob = new Blob([finalMasterContent], { type: 'application/vnd.apple.mpegurl' }); + const masterBlobUrl = URL.createObjectURL(masterBlob); + createdBlobs.push(masterBlobUrl); + + return { masterBlobUrl, allBlobs: createdBlobs }; + } catch (e) { + // Critical: Clean up any blobs created before the error + for (const blobUrl of createdBlobs) { + try { + URL.revokeObjectURL(blobUrl); + } catch { /* ignore cleanup errors */ } + } + console.error('[HLS Native] Recursive fetch failed', e); + throw e; + } + }; + + processMasterPlaylist(src).then((result) => { + video.src = result.masterBlobUrl; + extraBlobs = result.allBlobs; + }).catch((e) => { + console.warn('[HLS Native] Ad filtering failed, falling back to original source.', e); + onError?.('广告过滤失败，已回退到原始视频流'); + video.src = src; + }); + + } else { + video.src = src; + } } else { - console.error('[HLS] HLS not supported in this browser'); + console.error('[HLS] HLS not supported'); onError?.('当前浏览器不支持 HLS 视频播放'); } @@ -183,6 +330,7 @@ export function useHlsPlayer({ if (hls) { hls.destroy(); } + extraBlobs.forEach(url => URL.revokeObjectURL(url)); }; - }, [src, videoRef, autoPlay, onAutoPlayPrevented, onError]); + }, [src, videoRef, autoPlay, onAutoPlayPrevented, onError, isAdFilterEnabled, adFilterMode, adKeywords]); } diff --git a/components/player/hooks/usePlayerSettings.ts b/components/player/hooks/usePlayerSettings.ts index fe3f861..22208ac 100644 --- a/components/player/hooks/usePlayerSettings.ts +++ b/components/player/hooks/usePlayerSettings.ts @@ -1,7 +1,7 @@ 'use client'; import { useState, useEffect, useCallback } from 'react'; -import { settingsStore } from '@/lib/store/settings-store'; +import { settingsStore, AdFilterMode } from '@/lib/store/settings-store'; /** * Hook to access and update player settings from the settings store @@ -17,6 +17,9 @@ export function usePlayerSettings() { autoSkipOutro: stored.autoSkipOutro, skipOutroSeconds: stored.skipOutroSeconds, showModeIndicator: stored.showModeIndicator, + adFilter: stored.adFilter, + adFilterMode: stored.adFilterMode, + adKeywords: stored.adKeywords, }; }); @@ -31,6 +34,9 @@ export function usePlayerSettings() { autoSkipOutro: stored.autoSkipOutro, skipOutroSeconds: stored.skipOutroSeconds, showModeIndicator: stored.showModeIndicator, + adFilter: stored.adFilter, + adFilterMode: stored.adFilterMode, + adKeywords: stored.adKeywords, }); }); return unsubscribe; @@ -71,6 +77,18 @@ export function usePlayerSettings() { updateSetting('showModeIndicator', value); }, [updateSetting]); + const setAdFilter = useCallback((value: boolean) => { + updateSetting('adFilter', value); + }, [updateSetting]); + + const setAdFilterMode = useCallback((value: AdFilterMode) => { + updateSetting('adFilterMode', value); + }, [updateSetting]); + + const setAdKeywords = useCallback((value: string[]) => { + updateSetting('adKeywords', value); + }, [updateSetting]); + return { ...settings, setAutoNextEpisode, @@ -79,5 +97,8 @@ export function usePlayerSettings() { setAutoSkipOutro, setSkipOutroSeconds, setShowModeIndicator, + setAdFilter, + setAdFilterMode, + setAdKeywords, }; } diff --git a/components/ui/icons/utility-icons.tsx b/components/ui/icons/utility-icons.tsx index c154a7c..d0d554d 100644 --- a/components/ui/icons/utility-icons.tsx +++ b/components/ui/icons/utility-icons.tsx @@ -15,6 +15,14 @@ export const UtilityIcons = { ), + ShieldAlert: ({ className = "", size = 24 }: IconProps) => ( + + + + + + ), + Sparkles: ({ className = "", size = 24 }: IconProps) => ( diff --git a/lib/store/settings-store.ts b/lib/store/settings-store.ts index cc94c92..d2af77c 100644 --- a/lib/store/settings-store.ts +++ b/lib/store/settings-store.ts @@ -18,6 +18,7 @@ export type SortOption = | 'name-desc'; export type SearchDisplayMode = 'normal' | 'grouped'; +export type AdFilterMode = 'off' | 'keyword' | 'heuristic' | 'aggressive'; export interface AppSettings { sources: VideoSource[]; @@ -35,6 +36,9 @@ export interface AppSettings { autoSkipOutro: boolean; skipOutroSeconds: number; showModeIndicator: boolean; // Show '直连模式'/'代理模式' badge on player + adFilter: boolean; // Filter ad tags from m3u8 (legacy, kept for compatibility) + adFilterMode: AdFilterMode; // 'off' | 'keyword' | 'heuristic' | 'aggressive' + adKeywords: string[]; // Dynamically loaded ad keywords // Search & Display settings realtimeLatency: boolean; // Enable real-time latency ping updates searchDisplayMode: SearchDisplayMode; // 'normal' = individual cards, 'grouped' = group same-name videos @@ -87,51 +91,43 @@ function getEnvSubscriptions(customValue?: string): SourceSubscription[] { // Debugging helper // console.log("Environment Subscriptions:", getEnvSubscriptions()); +// Shared default settings factory to avoid code duplication +function getDefaultAppSettings(): AppSettings { + return { + sources: getDefaultSources(), + premiumSources: getDefaultPremiumSources(), + subscriptions: getEnvSubscriptions(), + sortBy: 'default', + searchHistory: true, + watchHistory: true, + passwordAccess: false, + accessPasswords: [], + autoNextEpisode: true, + autoSkipIntro: false, + skipIntroSeconds: 0, + autoSkipOutro: false, + skipOutroSeconds: 0, + showModeIndicator: false, + adFilter: false, + adFilterMode: 'heuristic', + adKeywords: [], + realtimeLatency: false, + searchDisplayMode: 'normal', + episodeReverseOrder: false, + }; +} + export const settingsStore = { getSettings(): AppSettings { + // SSR: Return defaults if (typeof window === 'undefined') { - return { - sources: getDefaultSources(), - premiumSources: getDefaultPremiumSources(), - subscriptions: getEnvSubscriptions(), - sortBy: 'default', - searchHistory: true, - watchHistory: true, - passwordAccess: false, - accessPasswords: [], - autoNextEpisode: true, - autoSkipIntro: false, - skipIntroSeconds: 0, - autoSkipOutro: false, - skipOutroSeconds: 0, - showModeIndicator: false, - realtimeLatency: false, - searchDisplayMode: 'normal', - episodeReverseOrder: false, - }; + return getDefaultAppSettings(); } + // Client: No stored settings, return defaults const stored = localStorage.getItem(SETTINGS_KEY); if (!stored) { - return { - sources: getDefaultSources(), - premiumSources: getDefaultPremiumSources(), - subscriptions: getEnvSubscriptions(), - sortBy: 'default', - searchHistory: true, - watchHistory: true, - passwordAccess: false, - accessPasswords: [], - autoNextEpisode: true, - autoSkipIntro: false, - skipIntroSeconds: 0, - autoSkipOutro: false, - skipOutroSeconds: 0, - showModeIndicator: false, - realtimeLatency: false, - searchDisplayMode: 'normal', - episodeReverseOrder: false, - }; + return getDefaultAppSettings(); } try { @@ -187,33 +183,16 @@ export const settingsStore = { autoSkipOutro: parsed.autoSkipOutro !== undefined ? parsed.autoSkipOutro : false, skipOutroSeconds: typeof parsed.skipOutroSeconds === 'number' ? parsed.skipOutroSeconds : 0, showModeIndicator: parsed.showModeIndicator !== undefined ? parsed.showModeIndicator : false, + adFilter: parsed.adFilter !== undefined ? parsed.adFilter : false, + adFilterMode: parsed.adFilterMode || 'heuristic', + adKeywords: Array.isArray(parsed.adKeywords) ? parsed.adKeywords : [], realtimeLatency: parsed.realtimeLatency !== undefined ? parsed.realtimeLatency : false, searchDisplayMode: parsed.searchDisplayMode === 'grouped' ? 'grouped' : 'normal', episodeReverseOrder: parsed.episodeReverseOrder !== undefined ? parsed.episodeReverseOrder : false, }; } catch { // Even if localStorage fails, we should return defaults + ENV subscriptions - const envSubscriptions = getEnvSubscriptions(); - - return { - sources: getDefaultSources(), - premiumSources: getDefaultPremiumSources(), - subscriptions: envSubscriptions, - sortBy: 'default', - searchHistory: true, - watchHistory: true, - passwordAccess: false, - accessPasswords: [], - autoNextEpisode: true, - autoSkipIntro: false, - skipIntroSeconds: 0, - autoSkipOutro: false, - skipOutroSeconds: 0, - showModeIndicator: false, - realtimeLatency: false, - searchDisplayMode: 'normal', - episodeReverseOrder: false, - }; + return getDefaultAppSettings(); } }, diff --git a/lib/utils/m3u8-ad-detector.ts b/lib/utils/m3u8-ad-detector.ts new file mode 100644 index 0000000..05912f5 --- /dev/null +++ b/lib/utils/m3u8-ad-detector.ts @@ -0,0 +1,265 @@ +/** + * Heuristic Ad Detection Module + * + * Provides block-based analysis for detecting ads in M3U8 playlists + * using filename pattern matching and other heuristics. + */ + +// Ad-related path keywords for scoring +export const AD_PATH_KEYWORDS = [ + 'advert', 'preroll', 'midroll', 'postroll', + 'dai', 'vast', 'ima', 'adjump', 'commercial', 'sponsor' +]; + +/** + * Represents a segment in the playlist + */ +interface Segment { + url: string; + duration: number; + lineIndex: number; +} + +/** + * Represents a block of segments between DISCONTINUITY markers + */ +interface Block { + segments: Segment[]; + startLineIndex: number; + endLineIndex: number; + hasCueTag: boolean; +} + +/** + * Pattern extracted from main content for comparison + */ +interface MainPattern { + filenameRegex: RegExp | null; + avgDuration: number; + commonPrefix: string; + pathPrefix: string; // Directory path prefix (e.g., "/20230907/73PWifvT/1392kb/hls/") +} + +/** + * Parse M3U8 content into blocks separated by DISCONTINUITY markers + */ +export function parseBlocks(lines: string[]): Block[] { + const blocks: Block[] = []; + let currentBlock: Block = { + segments: [], + startLineIndex: 0, + endLineIndex: 0, + hasCueTag: false + }; + + for (let i = 0; i < lines.length; i++) { + const line = lines[i].trim(); + + // Check for CUE tags + if (line.startsWith('#EXT-X-CUE-OUT') || line.startsWith('#EXT-X-CUE-IN')) { + currentBlock.hasCueTag = true; + } + + // DISCONTINUITY marks block boundary + if (line === '#EXT-X-DISCONTINUITY') { + if (currentBlock.segments.length > 0) { + currentBlock.endLineIndex = i - 1; + blocks.push(currentBlock); + } + currentBlock = { + segments: [], + startLineIndex: i + 1, + endLineIndex: 0, + hasCueTag: false + }; + continue; + } + + // Parse EXTINF and the following URL + if (line.startsWith('#EXTINF:')) { + const durationMatch = line.match(/#EXTINF:([\d.]+)/); + const duration = durationMatch ? parseFloat(durationMatch[1]) : 0; + + // Next line should be the URL + if (i + 1 < lines.length) { + const url = lines[i + 1].trim(); + if (url && !url.startsWith('#')) { + currentBlock.segments.push({ + url, + duration, + lineIndex: i + 1 + }); + } + } + } + } + + // Don't forget the last block + if (currentBlock.segments.length > 0) { + currentBlock.endLineIndex = lines.length - 1; + blocks.push(currentBlock); + } + + return blocks; +} + +/** + * Extract filename from URL (handles both relative and absolute URLs) + */ +function extractFilename(url: string): string { + try { + const path = url.includes('://') ? new URL(url).pathname : url; + const parts = path.split('/'); + return parts[parts.length - 1] || ''; + } catch { + return url.split('/').pop() || ''; + } +} + +/** + * Find common prefix among an array of strings + */ +function findCommonPrefix(strings: string[]): string { + if (!strings || strings.length < 2) return ''; + + let prefix = ''; + const first = strings[0]; + + for (let i = 0; i < first.length; i++) { + const char = first[i]; + if (strings.every(s => s[i] === char)) { + prefix += char; + } else { + break; + } + } + + return prefix; +} + +/** + * Extract path prefix (directory) from URL + * e.g., "/20230907/73PWifvT/1392kb/hls/" from "/20230907/73PWifvT/1392kb/hls/gFE6lwIk.ts" + */ +function extractPathPrefix(url: string): string { + try { + const path = url.includes('://') ? new URL(url).pathname : url; + const lastSlash = path.lastIndexOf('/'); + return lastSlash >= 0 ? path.substring(0, lastSlash + 1) : ''; + } catch { + const lastSlash = url.lastIndexOf('/'); + return lastSlash >= 0 ? url.substring(0, lastSlash + 1) : ''; + } +} + +/** + * Learn pattern from the largest block (assumed to be main content) + */ +export function learnMainPattern(blocks: Block[]): MainPattern { + // Find the largest block by segment count (likely main content) + const mainBlock = blocks.length > 0 ? blocks.reduce((largest, block) => + block.segments.length > largest.segments.length ? block : largest + ) : null; + + if (!mainBlock || mainBlock.segments.length === 0) { + return { filenameRegex: null, avgDuration: 0, commonPrefix: '', pathPrefix: '' }; + } + + // Extract filenames + const filenames = mainBlock.segments.map(s => extractFilename(s.url)); + + // Find common prefix + const commonPrefix = findCommonPrefix(filenames); + + // Calculate average duration + const totalDuration = mainBlock.segments.reduce((sum, s) => sum + s.duration, 0); + const avgDuration = totalDuration / mainBlock.segments.length; + + // Try to build a regex pattern from the filenames + // Common patterns: "0000001.ts", "seg-1.ts", "segment_001.ts" + let filenameRegex: RegExp | null = null; + if (commonPrefix.length >= 2) { + // Escape special regex characters in prefix + const escapedPrefix = commonPrefix.replace(/[.*+?^${}()|[\]\\]/g, '\\$&'); + filenameRegex = new RegExp(`^${escapedPrefix}`); + } + + // Extract path prefix (directory path without filename) + // e.g., "/20230907/73PWifvT/1392kb/hls/" from "/20230907/73PWifvT/1392kb/hls/gFE6lwIk.ts" + const firstUrl = mainBlock.segments[0].url; + const pathPrefix = extractPathPrefix(firstUrl); + + return { filenameRegex, avgDuration, commonPrefix, pathPrefix }; +} + +/** + * Score a block for ad likelihood based on heuristics + * Returns a score where higher = more likely to be an ad + */ +export function scoreBlock(block: Block, mainPattern: MainPattern, extraKeywords: string[] = []): number { + let score = 0; + + // If block has CUE tag, it's definitely an ad + if (block.hasCueTag) { + return 10; // Max score + } + + // Check path keywords (Built-in + Custom) + // We filter out very short custom keywords to avoid false positives in scoring + const safeExtraKeywords = extraKeywords.filter(k => k.length > 2); + const allKeywords = [...AD_PATH_KEYWORDS, ...safeExtraKeywords]; + + for (const segment of block.segments) { + const urlLower = segment.url.toLowerCase(); + for (const keyword of allKeywords) { + if (urlLower.includes(keyword.toLowerCase())) { + score += 2.5; + break; // Only count once per segment + } + } + } + + // Check filename pattern mismatch + if (mainPattern.filenameRegex) { + const mismatchCount = block.segments.filter(s => { + if (!mainPattern.filenameRegex) return false; // No pattern to compare against + const filename = extractFilename(s.url); + return !mainPattern.filenameRegex.test(filename); + }).length; + + if (mismatchCount === block.segments.length && block.segments.length > 0) { + score += 1.5; // All filenames differ from main pattern + } + } + + // **KEY FEATURE**: Check path prefix mismatch (e.g., different date/folder/bitrate) + // This is the most reliable indicator for ads that come from different CDN paths + if (mainPattern.pathPrefix && block.segments.length > 0) { + const pathMismatchCount = block.segments.filter(s => { + const segmentPathPrefix = extractPathPrefix(s.url); + return segmentPathPrefix !== mainPattern.pathPrefix; + }).length; + + if (pathMismatchCount === block.segments.length) { + // ALL segments have different path prefix - strong ad indicator + score += 5.0; + } + } + + return score; +} + +/** + * Threshold configuration + */ +export const THRESHOLDS = { + HIGH: 5.0, // Definitely an ad + LOW: 3.0 // Possibly an ad (for future "fuzzy" mode) +}; + +/** + * Determine if a block should be filtered based on its score + */ +export function shouldFilterBlock(score: number, threshold: number = THRESHOLDS.HIGH): boolean { + return score >= threshold; +} diff --git a/lib/utils/m3u8-utils.ts b/lib/utils/m3u8-utils.ts new file mode 100644 index 0000000..4ccac7b --- /dev/null +++ b/lib/utils/m3u8-utils.ts @@ -0,0 +1,157 @@ + +/** + * Utility functions for M3U8 playlist manipulation + */ + +import { parseBlocks, learnMainPattern, scoreBlock, shouldFilterBlock } from './m3u8-ad-detector'; + +/** + * Filters ads from specific M3U8 content using multiple detection strategies: + * 1. Keyword matching (configurable via env) + * 2. CUE-OUT/CUE-IN standard tags + * 3. Heuristic block analysis (filename patterns, ad path keywords) + * + * Also converts relative URLs to absolute URLs for Blob playback. + * + * @param content The raw M3U8 content string + * @param baseUrl The base URL of the M3U8 file (to resolve relative paths) + * @returns The filtered M3U8 content + */ +export type AdFilterMode = 'off' | 'keyword' | 'heuristic' | 'aggressive'; + +export function filterM3u8Ad(content: string, baseUrl: string, mode: AdFilterMode = 'heuristic', customKeywords: string[] = []): string { + if (!content) return ''; + + // Use keywords passed from AdKeywordsWrapper (already loaded from env/file) + const keywords = customKeywords; + + const basePath = baseUrl.substring(0, baseUrl.lastIndexOf('/') + 1); + let origin = ''; + try { + origin = new URL(baseUrl).origin; + } catch (e) { /* ignore */ } + + // 2. Global Scan: Check if any ad keywords exist in the content + const hasKeywordMatch = mode !== 'off' && keywords.some(k => content.includes(k)); + const hasCueTag = mode !== 'off' && (content.includes('#EXT-X-CUE-OUT') || content.includes('#EXT-X-CUE-IN')); + + // 3. Heuristic Analysis: If no explicit ad signals, use block-based detection + const lines = content.split(/\r?\n/); + let adLineIndices = new Set(); + + if (!hasCueTag && (mode === 'heuristic' || mode === 'aggressive')) { + // No obvious ad signals - run heuristic analysis + const blocks = parseBlocks(lines); + if (blocks.length > 1) { + const mainPattern = learnMainPattern(blocks); + for (const block of blocks) { + // Pass all keywords (including custom ones) to heuristic scorer + const score = scoreBlock(block, mainPattern, keywords); + const threshold = mode === 'aggressive' ? 3.0 : 5.0; + if (shouldFilterBlock(score, threshold)) { + // Mark all lines in this block for removal + for (const segment of block.segments) { + adLineIndices.add(segment.lineIndex); + adLineIndices.add(segment.lineIndex - 1); // EXTINF line + } + } + } + } + } + + const processedLines: string[] = []; + + // State machine for CUE-OUT/CUE-IN tracking + let insideCueAdBlock = false; + + for (let i = 0; i < lines.length; i++) { + const line = lines[i]; + const trimmedLine = line.trim(); + + // Skip lines marked by heuristic analysis + if (adLineIndices.has(i)) { + continue; + } + + // 3. CUE Tag Detection (SCTE-35 Standard) + // EXT-X-CUE-OUT marks start of ad, EXT-X-CUE-IN marks end + if (mode !== 'off' && trimmedLine.startsWith('#EXT-X-CUE-OUT')) { + insideCueAdBlock = true; + // Remove preceding DISCONTINUITY if present + if (processedLines.length > 0 && processedLines[processedLines.length - 1].trim() === '#EXT-X-DISCONTINUITY') { + processedLines.pop(); + } + continue; // Skip the CUE-OUT tag itself + } + + if (trimmedLine.startsWith('#EXT-X-CUE-IN')) { + insideCueAdBlock = false; + // Also skip the next line if it's a DISCONTINUITY (ad block ending marker) + if (i + 1 < lines.length && lines[i + 1].trim() === '#EXT-X-DISCONTINUITY') { + i++; // Skip the following DISCONTINUITY + } + continue; // Skip the CUE-IN tag itself + } + + // Skip all content inside CUE ad block + if (insideCueAdBlock) { + continue; + } + + // 4. Keyword-based Ad Detection & Backtrack (skip if no keywords configured) + if (keywords.length > 0 && hasKeywordMatch && keywords.some(keyword => trimmedLine.includes(keyword))) { + // Found Ad: Remove it and backtrack to remove associated metadata + while (processedLines.length > 0) { + const lastIndex = processedLines.length - 1; + const lastLine = processedLines[lastIndex].trim(); + + if (lastLine.startsWith('#EXTINF:') || lastLine === '#EXT-X-DISCONTINUITY') { + processedLines.pop(); + } else { + break; + } + } + continue; // Skip the ad line itself + } + + // 5. Discontinuity Handling (Conservative Mode) + // Keep all Discontinuity tags by default. + // They will ONLY be removed via backtracking when a confirmed ad segment is found. + // This prevents false positives on legitimate concatenated streams. + if (trimmedLine === '#EXT-X-DISCONTINUITY') { + processedLines.push(line); + continue; + } + + // 6. General Cleanup & URL Normalization + if (!trimmedLine || trimmedLine.startsWith('http') || trimmedLine.startsWith('blob:')) { + processedLines.push(line); + continue; + } + + if (trimmedLine.startsWith('#')) { + // Handle URI="..." in attributes (e.g. #EXT-X-KEY) + if (trimmedLine.includes('URI="')) { + processedLines.push(line.replace(/URI="([^"]+)"/g, (match, uri) => { + if (uri.startsWith('http')) return match; // Already absolute + if (uri.startsWith('/')) { + return `URI="${origin}${uri}"`; // Root-relative + } + return `URI="${basePath}${uri}"`; // Path-relative + })); + } else { + processedLines.push(line); + } + continue; + } + + // 7. Resolve Relative URLs (for Blob support) + if (trimmedLine.startsWith('/')) { + processedLines.push(origin ? `${origin}${trimmedLine}` : trimmedLine); + } else { + processedLines.push(`${basePath}${trimmedLine}`); + } + } + + return processedLines.join('\n'); +}