在Miniflux 2的基础上添加了，图片，视频，音频缓存。 点击收藏时，自动下载到vps，取消后删除。缓存的内容不会被自动清除。 ai万岁 新加的

• MEDIA_CACHE_DIRECTORY=/app/mediacache

可选：单个媒体文件大小上限（MB），0 或不设=不限制。视频大就别设太小

• MEDIA_CACHE_MAX_FILE_SIZE_MB=0

volumes:

• ./mediacache:/app/mediacache

When a user stars an entry, its images/video/audio are downloaded to the VPS disk in the background; when the star is removed, the cached media is deleted (with reference counting so files shared across starred entries survive). Two new config options:

MEDIA_CACHE_DIRECTORY — set to enable the feature (empty = disabled, a no-op). Files shard as

/<hash[:2]>/. MEDIA_CACHE_MAX_FILE_SIZE_MB — per-file cap; 0 = unlimited.

Important operational note: cached media is served only through the existing /proxy/ endpoint. So the cache is useful only when the media proxy is enabled for the relevant types

MEDIA_PROXY_MODE must not be none (http-only or all), and MEDIA_PROXY_RESOURCE_TYPES must include the types you want served (default is image only — add video,audio to protect those).

ai的说明： ExtractMediaURLs gates on the exact same mediaproxy.ShouldProxifyURLWithMimeType decision the rewriter uses, so what gets cached and what gets served can never diverge. A DB migration runs on next startup, creating the entry_media_cache table (with a url_hash reference index and ON DELETE CASCADE on entries). Wired star paths: UI toggle, API toggle + batch + import, Fever saved/unsaved, Google Reader starred/unstarred — all fire go mediacache.SyncEntryStarredState(...). Safety: downloads go through the existing fetcher (SSRF-protected), cache paths are traversal-checked, the HMAC proxy signature is preserved, and http.ServeContent handles Range requests for video/audio seeking. The status check now accepts only 200 OK (never 206, since no Range header is sent). Verification: go build ./... exit 0, go vet clean on touched packages, mediacache tests pass, non-DB suites pass. (The one failing config test — TestParseAdminPasswordFileOptionWithEmptyFile — is a pre-existing environmental issue on Windows, confirmed identical on a clean tree.)

Miniflux 2

Miniflux is a minimalist and opinionated feed reader. It's simple, fast, lightweight and super easy to install.

Official website: https://miniflux.app

Features

Feed Reader

• Supported feed formats: Atom 0.3/1.0, RSS 1.0/2.0, and JSON Feed 1.0/1.1.
• OPML file import/export and URL import.
• Supports multiple attachments (podcasts, videos, music, and images enclosures).
• Plays videos from YouTube directly inside Miniflux.
• Organizes articles using categories and bookmarks.
• Share individual articles publicly.
• Fetches website icons (favicons).
• Saves articles to third-party services.
• Provides full-text search (powered by Postgres).
• Available in 20 languages: Portuguese (Brazilian), Chinese (Simplified and Traditional), Dutch, English (US), Finnish, French, German, Greek, Hindi, Indonesian, Italian, Japanese, Polish, Romanian, Russian, Taiwanese POJ, Ukrainian, Spanish, and Turkish.

Privacy and Security

• Removes pixel trackers.
• Strips tracking parameters from URLs (e.g., utm_source, utm_medium, utm_campaign, fbclid, etc.).
• Retrieves original links when feeds are sourced from FeedBurner.
• Opens external links with attributes rel="noopener noreferrer" referrerpolicy="no-referrer" for improved security.
• Implements the HTTP header Referrer-Policy: no-referrer to prevent referrer leakage.
• Provides a media proxy to avoid tracking and resolve mixed content warnings when using HTTPS.
• Plays YouTube videos via the privacy-focused domain youtube-nocookie.com.
• Supports alternative YouTube video players such as Invidious.
• Blocks external JavaScript to prevent tracking and enhance security.
• Sanitizes external content before rendering it.
• Enforces a Content Security and a Trusted Types Policy to only application JavaScript and blocks inline scripts and styles.

Bot Protection Bypass Mechanisms

• Optionally disable HTTP/2 to mitigate fingerprinting.
• Allows configuration of a custom user agent.
• Supports adding custom cookies for specific use cases.
• Enables the use of proxies for enhanced privacy or bypassing restrictions.

Content Manipulation

• Fetches the original article and extracts only the relevant content using a local Readability parser.
• Allows custom scraper rules based on CSS selectors.
• Supports custom rewriting rules for content manipulation.
• Provides a regex filter to include or exclude articles based on specific patterns.
• Optionally permits self-signed or invalid certificates (disabled by default).
• Scrapes YouTube's website to retrieve video duration as read time or uses the YouTube API (disabled by default).

User Interface

• Optimized stylesheet for readability.
• Responsive design that adapts seamlessly to desktop, tablet, and mobile devices.
• Minimalistic and distraction-free user interface.
• No requirement to download an app from Apple App Store or Google Play Store.
• Can be added directly to the home screen for quick access.
• Supports a wide range of keyboard shortcuts for efficient navigation.
• Optional touch gesture support for navigation on mobile devices.
• Custom stylesheets and JavaScript to personalize the user interface to your preferences.
• Themes:
  • Light (Sans-Serif)
  • Light (Serif)
  • Dark (Sans-Serif)
  • Dark (Serif)
  • System (Sans-Serif) – Automatically switches between Dark and Light themes based on system preferences.
  • System (Serif)

Integrations

• 25+ integrations with third-party services: Apprise, Betula, Cubox, Discord, Espial, Instapaper, LinkAce, Linkding, LinkTaco, LinkWarden, Matrix, Notion, Ntfy, Nunux Keeper, Pinboard, Pushover, RainDrop, Readeck, Readwise Reader, RssBridge, Shaarli, Shiori, Slack, Telegram, Wallabag, etc.
• Bookmarklet for subscribing to websites directly from any web browser.
• Webhooks for real-time notifications or custom integrations.
• Compatibility with existing mobile applications using the Fever or Google Reader API.
• REST API with client libraries available in Go and Python.

Authentication

• Local username and password.
• Passkeys (WebAuthn).
• Google (OAuth2).
• Generic OpenID Connect.
• Reverse-Proxy authentication.

Technical Stuff

• Written in Go (Golang).
• Single binary compiled statically without dependency.
• Works only with PostgreSQL.
• Does not use any ORM or any complicated frameworks.
• Uses modern vanilla JavaScript only when necessary.
• All static files are bundled into the application binary using the Go embed package.
• Supports the Systemd sd_notify protocol for process monitoring.
• Configures HTTPS automatically with Let's Encrypt.
• Allows the use of custom SSL certificates.
• Supports HTTP/2 when TLS is enabled.
• Updates feeds in the background using an internal scheduler or a traditional cron job.
• Uses native lazy loading for images and iframes.
• Compatible only with modern browsers.
• Adheres to the Twelve-Factor App methodology.
• Provides official Debian/RPM packages and pre-built binaries.
• Publishes a Docker image to Docker Hub, GitHub Registry, and Quay.io Registry, with ARM and RISC-V architecture support.
• Uses a limited amount of third-party go dependencies
• Has a comprehensive testsuite, with both unit tests and integration tests.
• Only uses a couple of MB of memory and a negligible amount of CPU, even with several hundreds of feeds.
• Respects/sends Last-Modified, If-Modified-Since, If-None-Match, Cache-Control, Expires and ETags headers, and has a default polling interval of 1h.

Documentation

The Miniflux documentation is available here: https://miniflux.app/docs/ (Man page)

• Opinionated?
• Features
• Requirements
• Installation Instructions
• Upgrading to a New Version
• Configuration
• Command Line Usage
• User Interface Usage
• Keyboard Shortcuts
• Integration with External Services
• Rewrite and Scraper Rules
• API Reference
• Development
• Internationalization
• Frequently Asked Questions

Screenshots

Default theme:

Dark theme when using keyboard navigation:

Credits

• Authors: Frédéric Guillot - List of contributors
• Distributed under Apache 2.0 License