Large language models get a polished front end that can run fully offline: Open WebUI is the self-hosted front end of choice. It talks to local model runners, primarily Ollama, and to any OpenAI-compatible API, so LM Studio, vLLM, Groq, Mistral, OpenRouter, and cloud providers all plug into the same chat interface and can be mixed per conversation. RAG is built in: upload files to knowledge bases or reference them in chat with the # command, backed by a choice of nine vector databases (ChromaDB and PGVector officially maintained) and multiple extraction engines including Tika and Docling, with hybrid BM25-plus-vector search and cross-encoder reranking. Web search results from providers like SearXNG, Brave, and Tavily inject directly into conversations. Extensibility comes from Python tools and functions that run inside the chat, a Pipelines plugin framework, and native MCP support. Multi-user features include RBAC, SSO, and group permissions, and the instance itself exposes an OpenAI-compatible API your own apps can call.

Agents with tool use, deep research, a document editor, an IMAP/SMTP email client with AI triage, notes, tasks, and a CalDAV-synced calendar - Odysseus bundles all of it into one open-source, self-hosted AI workspace. It runs local models through Ollama, vLLM, or llama.cpp and cloud APIs like OpenAI and OpenRouter, with a hardware-aware Cookbook that scans your machine and recommends quantized models that fit. Persistent memory uses ChromaDB with hybrid vector-plus-keyword retrieval, web search runs through a bundled SearXNG instance, and agents can use MCP servers, files, and shell access with safety controls, plus custom skills and scheduled agent tasks. A blind Compare mode runs side-by-side model duels with identities hidden and accumulates Elo-style ratings from your votes, so model selection is based on your actual workloads rather than leaderboard claims. Deep research mode - adapted from the Tongyi DeepResearch approach - reads sources through SearXNG and produces cited reports, while the email client tags, summarizes, sets reminders, and drafts replies locally rather than through a third-party mail AI. The writing-first document editor adds AI edits, Markdown and HTML support, and version history. The stack is Python 3.11 with FastAPI, SQLite for state, and a vanilla JS frontend, licensed AGPL-3.0 with zero telemetry. Because agents can read email and execute commands, keep authentication enabled and never expose it as a public unauthenticated service.

Every major model provider behind one ChatGPT-style interface: LibreChat spans OpenAI, Anthropic, Google, Azure, AWS Bedrock, Vertex AI, Groq, Mistral, OpenRouter, DeepSeek, and any OpenAI-compatible endpoint including local Ollama. You can switch models mid-conversation and compare providers without changing tools. Its Agents framework builds no-code custom assistants with tool access via Model Context Protocol servers, file search over uploaded documents through an optional pgvector-backed RAG service, and a sandboxed Code Interpreter that executes Python, JavaScript, Go, C++, Java, PHP, and Rust. Artifacts render React components, HTML, and Mermaid diagrams directly in chat, and image generation works through DALL-E and other configured providers. Multi-user support is enterprise-grade, with OAuth, SAML, LDAP, and two-factor authentication, per-user conversation history in MongoDB, and Meilisearch-powered search across all messages and files, plus reusable presets, forkable threads, and persistent memory across conversations. The economics favor teams: instead of a ChatGPT Plus seat per person, everyone shares one instance billed per API token, with access to every provider rather than one - and providers see individual API calls, not your accumulated organizational knowledge. Deployment is Docker Compose; API keys and endpoints are configured through .env and librechat.yaml.

On your own server, TypeType runs a private video platform for YouTube, NicoNico, and BiliBili, keeping subscriptions, history, and playlists out of someone else's recommendation engine. Pick a service and search it. Browse channels, Shorts, podcasts, and a home feed. Speed and quality control. Cinema mode. Native picture-in-picture. Subtitles, alternate audio tracks, and frame-by-frame stepping. SponsorBlock segments get skipped, or just marked, by category. Subscriptions, playlists, watch later, favorites, and resume positions fill the library. Import a Google Takeout export or a PipePipe backup and a new instance never starts empty. Download video or audio to local or S3-compatible storage and collect it in the browser when it is ready. Sign in locally or through OIDC. An admin console runs the instance. Block a channel and it disappears from feeds, search, and related videos. Hide Shorts. Hide personalized recommendations. Hide the suggestions column. A native Android client for phones and tablets syncs with the same instance. An optional experimental recommender can build a personalized home feed, and each user chooses whether to turn it on. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Open the page, write a Markdown note, move on - Memos is a lightweight, self-hosted service built for quick capture. Instead of folders, notebooks, and titles, it presents a timeline: open the page, write a Markdown note, and move on. Notes support headings, code blocks with syntax highlighting, task lists, tables, and file attachments, with tags auto-extracted from #hashtags in the text. Each memo carries a visibility level, private, protected (logged-in users), or public, so one instance works as a personal log, a small team wiki, or a lightweight microblog. The backend is a single Go binary with a React frontend, around 50 MB of memory at runtime and a ~20 MB Docker image, so it fits comfortably on the smallest instance size with near-zero maintenance. SQLite is the default store, with MySQL and PostgreSQL supported for multi-user deployments needing more concurrency, and full REST and gRPC APIs - Connect RPC for browsers, gRPC-Gateway for external tools - make capture scriptable from CLIs, bots, and automation platforms. Fast full-text search spans all memos, pinned notes keep references handy, and a masonry view suits visual browsing. MIT-licensed with zero telemetry; content is stored as plain Markdown in a database you control, so notes remain readable, exportable, and free of proprietary formats.

Perplexity's answer-engine experience, self-hostable and open-source: Morphic searches the web and writes cited answers. Instead of returning a list of links, it searches the web, reads the sources, and generates a complete answer with inline numbered citations. The generative UI streams rich components, source cards with thumbnails, image grids, syntax-highlighted code, and LaTeX math, rather than plain markdown. Quick mode answers fast; Adaptive mode runs deeper multi-step research. Search backends are pluggable: the Docker Compose bundle ships with a private SearXNG instance so no search API key is required, and Tavily, Brave, and Exa are supported alternatives. LLM providers include OpenAI, Anthropic, Google, Ollama, and any OpenAI-compatible endpoint, with per-mode model mapping - fast, cheap models for quick searches, stronger models for adaptive research, tuning the cost-quality trade-off per query type. An inspector panel exposes tool execution during multi-step research, and AI-suggested follow-up questions keep an investigation moving. Chat history persists in PostgreSQL, results are shareable by URL, file uploads feed context into queries, and optional Supabase authentication adds multi-user or guest access. Because the default search path is your private SearXNG instance, research topics never hit a commercial search API - and with local Ollama models the marginal cost of a query approaches zero. Built with Next.js, TypeScript, and the Vercel AI SDK under Apache 2.0.

Redlib lets you read Reddit without an account, ads, or trackers, through a clean page served from a machine you control. Open any subreddit, user profile, or comment thread and the server fetches the content and relays every image and video itself, so your browser never talks to Reddit directly. Pages need no JavaScript, unless you opt in to HLS video playback with sound. Choose card, clean, or compact layouts, a wide interface, and one of more than a dozen themes, from Dracula and Nord to Gruvbox and Tokyo Night. Post and comment sorting, a search box scoped to the current subreddit, a wiki tab, and subreddit filters help you find and tame your feed. Subscriptions build a personal home page with no login. Spoilers and NSFW previews can be blurred or hidden, and an SFW-only mode blocks mature content for the whole instance. Optional RSS feeds let a feed reader follow subreddits. Links to removed posts open in a configurable undelete viewer. Operators set defaults through environment variables, add an instance banner, discourage search indexing, and route outbound traffic through an HTTP or SOCKS proxy. Settings live in a browser cookie and can be exported or imported between devices. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0 licensed.

GoatCounter delivers meaningful web traffic insights — pageviews, referrers, browsers, screen sizes, country-level geolocation — without setting a single cookie, without collecting personal data, and without forcing GDPR consent banners on your visitors. Written entirely in Go and distributed as a single compiled binary consuming roughly 25MB of RAM, it adds just 3.5KB to your pages via the tracking script, with a JavaScript-free tracking pixel alternative for sites that avoid scripts entirely, plus backend middleware integration and log file import for server-side collection. The dashboard displays pageview counts per path with hourly resolution, referrer sources grouped by domain with full URL on hover, browser and OS version breakdowns, screen size distributions, and country-level location data derived from IP addresses that are immediately discarded after geolocation. Campaign tracking supports UTM parameters and custom data attributes. A public stats option exposes your dashboard at a shareable URL for build-in-public transparency. SQLite serves as the default database requiring zero administration, while PostgreSQL handles higher-traffic deployments with multi-site setups. Built-in ACME and TLS certificate management eliminates reverse proxy requirements for HTTPS — no Nginx or Caddy needed. The REST API provides programmatic access to all analytics data. Deploy as a single binary, via Docker with the official arp242/goatcounter image, or through native packages. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. EUPL-1.2 licensed.

Running a VPN for family, friends, or a small team usually means babysitting servers one SSH session at a time, and Mistgate replaces that with one panel that keeps the whole fleet configured. Add a server from the browser by typing its address and root password: the panel pins the host key, checks the host, opens the ports, and installs the signed node agent for you. Each person gets one subscription link that their app turns into the server list, or a personal AmneziaVPN key per device, so a lost phone can be revoked without touching anyone else. Groups decide which protocol profiles a person receives, while per-user traffic quotas, expiry dates, device limits, and per-device speed limits keep sharing fair. Hysteria2 brings QUIC-based speed with port hopping, obfuscation, and certificate options; VLESS REALITY and AmneziaWG 2.0 and 3.1 cover networks where one protocol gets blocked, and Cloudflare WARP can serve as the exit for all three. Every five minutes the panel connects to each server the way a real client app would, a doctor on every node runs host checks with safe fixes, and alerts reach you on Telegram before your users notice. The public address shows strangers a decoy site while the admin hides behind a secret path, and encrypted backups go to your Cloudflare R2 bucket on a schedule. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0 licensed.

The utilities engineers otherwise scatter across a dozen ad-laden websites - 80+ of them - live together in IT-Tools, one fast, polished web app. Crypto covers JWT decoding, MD5 through SHA-512 hashing, HMAC and bcrypt generation, RSA key pairs, and password strength analysis. Converters handle JSON to CSV, YAML, and TOML, Base64 files, URL encoding, HTML entities, color formats, and Docker run commands to Compose files. Generators produce UUIDv4, ULID, BIP39 mnemonics, QR codes (including Wi-Fi QR), and tokens; text tools include a regex tester, diff viewer, slug and case converters; web utilities parse URLs and user agents, look up HTTP status codes and MIME types, and inspect Open Graph metadata; plus a cron parser, chmod calculator, and more. The privacy argument is the point: JWTs contain user IDs, hashes derive from passwords, JSON dumps hold PII - exactly the inputs you least want a third-party utility site to log. IT-Tools is a frontend-only static bundle (Vue/TypeScript, GPL-3.0, 39k+ GitHub stars) served by Nginx in one container, so everything runs client-side on your infrastructure with nothing transmitted anywhere. New tools ship roughly monthly, and a scaffolding script makes adding custom ones straightforward.

Roughly $1M of development work, open-sourced: Maybe Finance began as a $249/year commercial personal finance product before the company released it all. It aggregates bank accounts, credit cards, loans, investments, crypto, and real estate into a single net worth dashboard with historical trend charts - replacing the spreadsheet that usually glues a whole portfolio together. Transactions are categorized and tagged with rules, with merchant tracking and search across imported or synced activity; budgets track spending by category against plan; and the investment view follows holdings, cost basis, and returns across brokerage accounts. Multi-currency support converts accounts held in different currencies into a single reporting currency, bank synchronization works through Plaid where supported, and manual CSV import covers any institution. An optional AI assistant answers questions grounded in your own financial data. Because the app was built as a paid product with professional design before being open-sourced, its interface quality exceeds most community finance tools - and self-hosting means your balances and transactions are not monetized by a free app or gated behind an annual subscription. The stack is Ruby on Rails with Hotwire on PostgreSQL, licensed AGPL-3.0 and deployed via Docker. The original repository is archived; development continues in the community fork Sure, compatible with the same self-hosted setup.

PrivateBin shares text and files with true zero-knowledge encryption: the server stores only ciphertext it cannot decrypt because the key exists exclusively in the URL fragment that never reaches the server. With over 8,400 GitHub stars and 500,000+ Docker pulls, it has become the standard for privacy-respecting paste sharing where not even the hosting operator can read stored content. All encryption uses 256-bit AES in Galois Counter mode performed entirely client-side in the browser. Burn-after-reading mode destroys pastes on first view, while configurable expiration timers automatically purge content after periods ranging from five minutes to forever. Optional password protection adds a second encryption layer requiring recipients to enter a passphrase before decryption occurs. Discussion threads enable anonymous or nickname-based commenting on pastes with IP-based identicon avatars for visual identity. Markdown rendering with live preview formats pastes as styled HTML, and syntax highlighting through prettify.js supports dozens of programming languages across four color themes. File upload support enables sharing images, PDFs, and media files with in-browser preview. QR code generation converts paste URLs for mobile device transfer. Customizable templates include Bootstrap 5, classic Bootstrap CSS, and Darkstrap themes. Storage backends support filesystem, MySQL, PostgreSQL, SQLite, Amazon S3, and Google Cloud Storage. Deploy via Docker with the privatebin/nginx-fpm-alpine image on port 8080. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. zlib licensed.

Web analytics tools ignore native mobile, desktop, and game apps; Aptabase was built for exactly those. If Firebase Analytics would force a privacy-policy footnote you don't want to write, this is the alternative - session-based metrics with no cookies, no IDFA or GAID, no device fingerprinting, and a daily-rotated salt that makes cross-day re-identification mathematically impossible. That design means GDPR, CCPA, and PECR compliance out of the box and "Data Not Collected" App Store privacy labels without ATT prompts. The SDK coverage is the widest in its category: eleven first-party libraries spanning Swift, Kotlin, Flutter, React Native, Tauri, Electron, .NET MAUI, NativeScript, Unity, Unreal Engine, and JavaScript for web - each MIT-licensed, following platform conventions, and accepting a custom host parameter that points at your instance. Integration is minutes: initialize with an app key, call trackEvent with optional properties, and the dashboard shows sessions, events, app versions, OS breakdowns, and country-level geography. The self-hosted stack is a .NET server over PostgreSQL for metadata and ClickHouse for high-volume event ingestion, giving cloud-parity features under an AGPL license. For indie iOS/Android apps, Electron and Tauri tools, and Unity or Unreal games, it replaces Firebase without the Google entanglement.

With over 48,000 GitHub stars and monthly releases since March 2023, LocalAI is the self-hosted AI engine that replaces every OpenAI endpoint with a single Docker container running on your own infrastructure — serving chat completions, image generation, text-to-speech, speech-to-text, embeddings, vision, video generation, and function calling through identical API schemas that require zero application code changes. The composable backend architecture isolates each inference engine as a separate gRPC service running in its own OCI container, so llama.cpp, vLLM, SGLang, transformers, whisper.cpp, diffusers, MLX, Stable Diffusion, and Flux install on demand without touching the core, can run on separate machines, and a fault in one never affects others. Hardware acceleration spans NVIDIA CUDA 12 and 13, AMD ROCm, Intel oneAPI/SYCL, Apple Silicon Metal, Vulkan, and NVIDIA Jetson L4T — or runs entirely on CPU without any GPU. Built-in AI agents support autonomous tool use, retrieval-augmented generation, Model Context Protocol integration, and skill-based workflows directly in the web interface. The model gallery provides curated YAML configuration files for hundreds of models that install with a single command, while P2P federated inference distributes model shards across multiple machines for running models larger than any single node's memory. Multi-user API key authentication with quotas and role-based access enables team deployments. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Your cycle history is among the most personal data you will ever record, and Ovumcy keeps it on a server you control instead of a tracker's cloud. Log period days, flow intensity, symptoms, and free-form notes in seconds from any browser, and define your own custom symptoms when the preset list misses something you feel. The calendar view lays out your month with predicted period, ovulation, and fertile-window days marked, while statistics views surface cycle length trends and patterns across months. Predictions come from your own logged history, refined when temperature or cervical-mucus entries sharpen the estimate. Reminders reach you three ways: a dashboard banner before an expected period, a webhook ping to your own ntfy or Gotify server, or a private read-only calendar subscription that drops estimates into the calendar app you already use. Install it to your phone home screen and it opens like a native app, in English, Russian, Spanish, French, German, or Italian. Export the full history as CSV or JSON whenever you want a backup or a change of tools. Optional OIDC sign-in and TOTP two-factor authentication protect access, and the app makes no analytics or telemetry calls at all. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0 licensed.

Grammar, punctuation, and style errors a dictionary lookup can't see: LanguageTool is open-source proofreading powered by a Java rule engine covering English, German, Spanish, French, Portuguese, Dutch, and 25+ other languages. Self-hosting the HTTP server is how you get Grammarly-class checking without sending every sentence you write to a third party - a real concern when the text being proofread is confidential email, legal drafts, or unreleased documentation. Your instance exposes the standard /v2/check API, so the official ecosystem plugs straight in: browser extensions for Chrome and Firefox accept a custom server URL, and integrations exist for VS Code, LibreOffice, Obsidian, Vim, Emacs, and many editors. Notably, self-hosting restores free browser-extension checking that the hosted service moved behind a premium subscription - your server, no character limits, no paywall. Detection quality is tunable: optional n-gram datasets (multi-gigabyte language models for en, de, es, fr, nl) teach the engine word-order and confusion-pair errors like there/their and brakes/breaks, and a fastText model improves automatic language identification. Everything runs offline once models are downloaded. The core is LGPL, the API is documented with Swagger, and rules are community- maintained and constantly expanding.

The ad-riddled "free online tools" sites people paste sensitive text into and upload confidential PDFs to - OmniTools replaces that whole ecosystem with one self-hosted app. It bundles 50+ utilities behind one clean React/TypeScript interface: image tools (resize, convert, crop, edit), video and audio tools (trim, reverse, convert), PDF tools (split, merge, edit), text and list utilities (case converters, formatters, shufflers), plus date/time, math, and data-format helpers for JSON, CSV, and XML. The architectural decision that makes it trustworthy is that all file processing happens entirely client-side in the browser - the server only serves static assets, and nothing you process ever leaves your device. That design has a pleasant side effect: the host needs almost no resources (people run it on a Raspberry Pi Zero), because your browser does the work while the server just delivers files. The Docker image is a remarkable 28 MB, making it one of the fastest apps to deploy and cheapest to keep running. There are no ads, no tracking, no accounts, and no upload limits. With multi-language support and an MIT license, it works equally well as a personal toolbox or a team-wide internal utility portal - one URL that replaces a bookmark folder full of questionable converters. Actively developed with 50 contributors and 9,500+ GitHub stars.

With over 8,000 GitHub stars and continuous development since 2020, LibrePhotos delivers the core intelligence of Google Photos — face recognition, object detection, semantic search, and automatic album generation — entirely on your own hardware without sending a single photo to a third-party server. The Django 5 backend processes uploaded media through a machine learning pipeline that runs face detection via the face_recognition library, clusters identified faces using scikit-learn and HDBSCAN, generates image captions through BLIP and Moondream 2, and classifies scenes using Places365 or Google's SigLIP 2 vision-language model with zero-shot classification against 900+ real-world tags. Semantic search lets you find photos by natural language queries like "sunset at the beach" without manual tagging, while metadata search filters by person, camera, lens, file type, and filesystem path. The React 18 frontend built with Vite presents a timeline view, fullscreen lightbox with slideshow mode, photo detail sidebar showing location and people, and a folder navigation view with breadcrumb paths. RAW files from any camera are converted via ImageMagick, videos processed through FFmpeg, and Live Photos paired with their RAW+JPEG counterparts as unified entries. Public album sharing via link provides fine-grained privacy controls, and duplicate detection uses perceptual hashing to identify near-identical images. Deployment runs as a single unified Docker container or via Docker Compose with Kubernetes manifests also available. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.