Sapphire is a self-hosted AI companion you can shape into a personal assistant, smart-home voice, or research partner that remembers conversations and acts through the tools you choose. Choose from twelve built-in personas or build your own, with each persona bundling its prompt, voice, tools, and model, and share personas, prompts, and toolsets as portable JSON files. Talk hands-free with wake word, speech-to-text, streaming text-to-speech, and barge-in, or type in chats that each keep their own prompt, tools, voice, and memory scope. Recall past details through semantic memory search, knowledge folders with file upload and folder watching, entity records for people and places, and hierarchical goals with a timestamped progress journal. Connect Discord, Telegram, email, Google Calendar, Home Assistant, Twilio phone calls, webcams, and image generation, then extend everything through one-click plugins, toolsets, and both an MCP server and an MCP client. Schedule heartbeats, morning greetings, alarms, background agents, daemons, and webhooks so the assistant checks in, follows goals overnight, and reports back when work finishes. Keep control with a privacy vault for encrypted chats, plugin signing and trust levels, in-app backup and restore, and a dashboard showing token metrics and system controls. 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.

Project AIRI is the most popular open-source AI companion platform — a self-hosted recreation of Neuro-sama that brings AI-powered virtual characters into your world across web, desktop, and mobile. The system renders Live2D, Spine, and VRM 3D character models with auto-blink, eye tracking, and lip-sync driven by real-time voice synthesis, while the xsAI abstraction layer connects to 40+ LLM providers including OpenAI GPT-4, Anthropic Claude, Google Gemini, DeepSeek, and local models via Ollama and OpenRouter. Built from day one on WebGPU, WebAudio, Web Workers, WebAssembly, and WebSocket technologies, the browser version runs entirely client-side with PWA offline support while the server runtime enables persistent memory via PostgreSQL with pgvector embeddings and DuckDB WASM for client-side storage. The Minecraft agent plays autonomously using mineflayer with pathfinding, and a Factorio integration provides cooperative gameplay. Social integrations deploy your companion as a Discord bot joining voice channels, a Telegram bot, and a Twitter/X agent posting and replying autonomously. The desktop Stage Tamagotchi app provides an always-on-screen companion for Windows and macOS, while Stage Pocket brings the experience to mobile. Voice features include client-side speech recognition via VAD, multiple TTS providers including ElevenLabs, and screen vision capabilities. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

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.

A private ChatGPT built with Next.js: Lobe Chat is the open-source AI chat interface teams self-host instead. Its main advantage is provider breadth: one interface connects to 40+ model providers, including OpenAI, Anthropic Claude, Google Gemini, Mistral, Groq, AWS Bedrock, Azure, and local models served through Ollama, so you can switch models per conversation and compare outputs. It handles multi-modal work: image recognition, image generation, text-to-speech, and speech-to-text. A plugin system based on function calling and the Model Context Protocol (MCP) adds external tools like web search and code execution. Run it in standalone mode as a single container with settings in browser storage, or in database mode with PostgreSQL and S3-compatible storage for persistent history, multi-user auth, and RAG knowledge bases built from uploaded documents with pgvector retrieval. Because tools arrive through function calling and MCP rather than a proprietary plugin format, custom internal tools can be exposed to the assistant with a standard server over STDIO or HTTP. Hundreds of pre-configured assistant roles import from the community marketplace. For teams the cost model matters: provider API keys billed per token typically undercut a ChatGPT Plus seat per person, and self-hosting keeps API keys, uploaded files, embeddings, and conversation history entirely on your own server.

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.

An 8MB Go binary that boots in under one second, uses less than 10MB of RAM, yet delivers full AI agent capabilities across 16+ chat platforms simultaneously. PicoClaw connects to Telegram, Discord, Matrix, IRC, Slack, WeCom, DingTalk, WeChat, LINE, and QQ while supporting LLM providers spanning OpenAI, Anthropic, Gemini, DeepSeek, AWS Bedrock, Azure, and local models via Ollama. Native Model Context Protocol support enables standardized tool integration, and the built-in smart routing engine directs simple queries to lightweight models to reduce API costs while sending complex tasks to capable models. Tool capabilities include secure shell execution, filesystem access, web search, cron scheduling for recurring tasks, and sub-agent spawning with status tracking. Gateway mode transforms PicoClaw into a full AI backend with REST API endpoints accessible from any client. The Skills system loads hierarchical behavior definitions from SKILL.md files, enabling customizable agent personalities and workflows. Compiles for x86_64, ARM64, ARMv7, RISC-V, MIPS, and LoongArch, making it deployable on hardware as cheap as a $10 Sipeed LicheeRV Nano. Achieved nearly 30,000 stars within six months of its February 2026 release. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

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.

With 26,000 GitHub stars and consistent victories over ElevenLabs in blind evaluations, Chatterbox delivers state-of-the-art text-to-speech with zero-shot voice cloning requiring only 5 seconds of reference audio. The model family spans three architectures: Chatterbox Multilingual V3 (500M parameters, 23+ languages including Arabic, Chinese, Japanese, Korean, Hindi, French, German, Spanish, and Portuguese), Chatterbox-Turbo (350M parameters optimized for voice agents with a single-step distilled decoder achieving ~200ms time-to-first-speech), and Chatterbox-Nano (110M parameters running 3x faster than realtime on 8 CPU cores for edge deployment). Unique among open-source TTS systems, Chatterbox introduces emotion exaggeration control — adjusting intensity from monotone to dramatically expressive via a single parameter — and native paralinguistic tagging where tokens like [laugh], [cough], [chuckle], and [gasp] inject natural vocal reactions inline without post-processing. The alignment-informed inference pipeline eliminates hallucinations and repetition artifacts common in autoregressive TTS. Built-in PerTh neural watermarking embeds imperceptible forensic identifiers in generated audio for provenance tracking. Trained on 500,000 hours of cleaned speech data across all supported languages. Voice conversion scripts enable transforming existing audio into any cloned voice. Deploy via pip install with PyTorch, serve through Gradio interfaces or custom FastAPI endpoints, and expose via HTTP streaming or WebSocket for sub-200ms conversational applications. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Kokoro-FastAPI turns text into natural-sounding speech across eight languages by serving the 82-million-parameter Kokoro-82M model through an OpenAI-compatible REST API, so any existing OpenAI SDK client can generate audio by just changing the base URL. With over 5,300 GitHub stars since December 2024, the fully Dockerized FastAPI server covers American English, British English, Spanish, French, Hindi, Italian, Japanese, Brazilian Portuguese, and Mandarin Chinese with language-specific phoneme processing. Inline voice mixing blends multiple profiles using weighted ratios like af_bella(2)+af_heart(1), automatically normalizing weights and caching combined voicepacks as PyTorch tensor files for reuse. Audio streams in real time over HTTP with configurable chunk sizes, or generates complete files in MP3, WAV, OPUS, FLAC, AAC, or PCM formats with speed control from 0.25x to 4.0x. Per-word timestamped captions with speaker-tagged voice labels enable subtitle generation for podcasts, audiobooks, and accessibility workflows. Pre-built Docker images support NVIDIA GPU acceleration via CUDA, experimental AMD GPU inference via ROCm, and CPU-only deployment on linux/amd64 and linux/arm64 architectures, with Apple Silicon MPS support available through direct UV execution. The integrated web interface at port 8880 provides browser-based speech generation, while the Swagger UI at /docs exposes the full API reference. Debug endpoints report system statistics for monitoring inference load. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache 2.0 licensed.

Your house already has lights, locks, sensors, thermostats, and speakers from many brands, and Home Assistant gives them one local control room with no vendor cloud in the loop. To close the shutters when the outdoor temperature holds above 25 °C, or to be alerted when the garage door is left open, build the routine in the visual editor: When, And if, Then do. Scenes, scripts, and blueprints package it for reuse, and run traces show what it did. More than 1,500 integrations cover Zigbee, Z-Wave, Matter, Thread, MQTT, ESPHome, Bluetooth, Hue, and Sonos, and network discovery offers known devices for one-step setup. To see the house at a glance, drag tile, thermostat, gauge, graph, and map cards into a dashboard, then put it on a TV with Home Assistant Cast. The Energy dashboard answers where the power goes, drawing solar, grid, battery, gas, and water as one distribution diagram with cost totals and per-appliance tracking. To talk to the house, use Assist: natural-language commands, wake words, and custom sentences run locally or through Home Assistant Cloud. The map panel shows who is home, plotting people against zones such as Home, Work, and School from the iOS and Android companion apps. Areas and labels keep large setups tidy, and multi-factor authentication guards each account. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.

With over 20,000 GitHub stars and adoption by companies building everything from telehealth platforms to AI voice agents, LiveKit is the most widely deployed open-source real-time communication server available. The Go-based Selective Forwarding Unit handles hundreds of concurrent participants per node with adaptive bitrate streaming, simulcast layers, SVC codec support for VP9 and AV1, and end-to-end encryption. Client SDKs span JavaScript, Swift, Kotlin, Flutter, React Native, Rust, Python, Unity, and ESP32 embedded devices, while server-side APIs cover Node.js, Go, Ruby, Java, Python, Rust, PHP, and .NET. The Agents framework enables building AI-powered voice and video applications — real-time speech-to-text, LLM-driven conversations, and computer vision pipelines — running as server-side participants in any room. Egress records sessions to S3-compatible storage or streams to RTMP endpoints, while Ingress pulls external feeds from OBS via RTMP, WHIP, or SRT into LiveKit rooms. The SIP bridge connects traditional telephony to WebRTC rooms for hybrid conferencing. JWT-based authentication, webhook notifications, room-level moderation APIs, and selective subscription give operators granular control. Deploy as a single binary for development, Docker Compose for production single-node, or Kubernetes with the official Helm chart for distributed multi-region clusters using Redis for state coordination. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache 2.0 licensed.

Build a voice AI agent that answers calls, qualifies leads, books appointments, and transfers to a human when needed, all from a drag-and-drop workflow builder in your browser. Dograh ships as a Docker Compose stack (API, web UI, Postgres, Redis, MinIO) that you self-host on any Linux server with automatic HTTPS provisioning via Let's Encrypt. The visual workflow builder lets you design multi-turn conversation flows by connecting nodes for greetings, intent classification, tool calls, and handoffs; describe your use case in plain English and the platform generates the LLM prompt and node graph for you. Connect your own speech-to-text, LLM, and text-to-speech providers (OpenAI, Anthropic, Gemini, ElevenLabs, Deepgram, local Whisper, Kokoro, or any OpenAI-compatible endpoint) or use the built-in Speech-to-Speech mode with Gemini Flash Live and GPT-Realtime-2 for sub-200ms latency. Telephony plugs in through Twilio, Vonage, Vobiz, or Cloudonix for inbound and outbound calling, with live agent transfer when the conversation needs a human. Webhook tool calls connect to Salesforce, HubSpot, Google Calendar, Cal.com, or any REST API without writing orchestration code. The ClonedVoice feature mixes real human voice recordings for high-frequency phrases with neural TTS fallback for dynamic content, cutting costs while improving caller trust. A built-in MCP server lets AI coding agents like Claude Code or Cursor design, test, and edit workflows through natural language. Deploy on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. BSD 2-Clause licensed.

VoicEra lets you run your own voice AI phone platform, where callers talk to agents you configure and every recording and transcript stays on infrastructure you control. Build telephony agents that answer inbound calls or run outbound campaigns, choose the speech recognition, language model, and speech synthesis behind each one, and swap providers without rebuilding the platform. Connect phone numbers through telephony carriers, route live audio through the real time runtime, and let workers execute campaign calls in the background while you watch progress from the operator dashboard. Review call logs with recordings and transcripts after each conversation, including provider recordings when carriers supply them. Serve callers in Indian and low resource languages by combining Bhashini, AI4Bharat, and other Indic models, switch languages mid call, or run speech and language models locally through the optional model server gateway. Add new AI or telephony providers behind stable registry interfaces when a better model or carrier appears, authenticate operators, and keep media in your own object storage and database. Test agents in the browser before any phone number is involved, then scale the same configuration to real calls. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.

Your front desk phone rings at midnight and nobody sane is awake; QuickVoice answers it with an AI agent you configured, on servers you own, with call recordings you control. Build agents in the web console: write the prompt, pick speech-to-text, language model, and voice, set the first message, then link a real phone number from Twilio or Telnyx through LiveKit SIP trunks. Inbound callers get a receptionist that answers questions from your uploaded knowledge base, PDFs and web pages chunked and embedded for live retrieval, while outbound mode fires single quick calls or batch campaigns for reminders and follow-ups. Agents can invoke allowlisted MCP tools mid-call, so checking an order status or booking a slot happens inside the conversation instead of after it. Every call lands in the log with transcript, recording, and metadata, searchable from the analytics dashboard, and privacy controls redact personal data by default, can suppress recordings entirely, and sweep old transcripts on a retention schedule. Multi-language voices, per-agent runtime overrides for campaigns, organization accounts with API keys and audit logs, and Stripe billing hooks round out a platform built for teams who would rather inspect their voice stack than rent a black box. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Once a voice agent is imported or built, Lampline connects it to a phone line, sends an AI caller to grade real calls, and keeps it all on your own server. Start by importing a Dialogflow CX or CX Agent Studio export along with a report of anything to check, or build fresh in a drag-and-drop builder that mixes LLM steps with scripted pages for exact lines and form fields. Next, pick one of 20 model providers, including Claude, OpenAI, Gemini, Bedrock, Groq, and DeepSeek, or run Ollama with local speech and no API keys. Test by chat or microphone while each step lights up, and let agents call HTTP and MCP tools and save records in collections of their own. Then connect callers through Twilio, Telnyx, or any SIP trunk, PBX, or session border controller via jambonz, with transfers to a person that carry the call's details as SIP headers. Finally, evals phone the agent over a real line, where an AI Test Caller brings background noise, accents, and interruptions, grades each transcript check by check, and reports typical and p95 response times as callers hear them. Each agent stays a folder of YAML and Markdown that you can review in git and lint or test with the vk CLI. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.