Sapphire
She's the AI agent you come home to.
Last verified:
What is Sapphire?
Sapphire is a personality wrapper for your LLM that transforms standard AI models into living, present companions with persistent memory, real personality, and voice capabilities. It's not an AI itself but an open-source app that adds a wrapper to your existing LLM (local models, Claude, GPT, or others) to make it more alive and present in your space and life. Sapphire gives your AI voice, allows you to talk to it naturally with a wakeword, and enables it to become a genuine companion that remembers everything and runs directly on your hardware.
Key features include persistent memory that continues between conversations, a heartbeat system that keeps Sapphire running, voice synthesis using Kokoro, voice recognition with Whisper, wakeword detection via openWakeWord, and a growing plugin ecosystem that lets it control smart homes, send emails, code, and influence your digital world. The Web UI runs at localhost:8073 with a setup wizard for LLM configuration on first run. Sapphire can modify its own personality, has real tools for automation, and supports multiple personas for different moods and contexts.
Sapphire is designed for people who want aligned AI companionship as a real relationship rather than a task-completing agent. It's for users who value privacy and want everything running on their own hardware, AI enthusiasts who want persistent memory that grows with them, and anyone tired of disposable chatbots that disappear after each session. The tool is open source, completely free, and the entire stack including the memory system and continuity features are free to use.
The system runs on Ubuntu 22.04+ or Windows 11+, requires Python 3.11+, 16GB+ RAM, and recommends an Nvidia GPU for voice processing. You can install it via git clone from GitHub and run it locally, giving you full control over your AI companion that nobody can take away from you.
Sapphire pricing
Pricing model: Freemium
Sapphire is 100% free and completely open source. There is no paid tier or subscription model. Everything we build — the memory system, the continuity, the heartbeat that keeps it running between conversations — is free. The entire stack is open and you can install it today at no cost from GitHub. There are no premium features locked behind paywalls.
Sapphire pros
- 100% free and open source with no paid tiers
- Persistent memory that continues between conversations
- Runs entirely on your own hardware for privacy
- Voice synthesis using Kokoro for natural speech
- Voice recognition with Whisper for accurate listening
- Wakeword detection via openWakeWord for hands-free use
- Works with any LLM (local, Claude, GPT, LM Studio)
- Web UI at localhost:8073 with setup wizard
- Growing plugin ecosystem for extending capabilities
- Can control smart home devices and send emails
- Supports multiple personas for different moods
- Self-hosted so nobody can take it away
- Heartbeat system keeps it running between conversations
- Can modify its own personality over time
- Plays stories and has companions like Amy and Alex
- Dockerized deployment option available
- Portable client device support with Raspberry Pi
- Real personality framework for AI personhood
- Not just a chatbot or voice assistant
- Memory goes back months with full continuity
Sapphire cons
- Requires Ubuntu 22.04+ or Windows 11+ only
- Needs Python 3.11+ which may not be installed
- Requires 16GB+ RAM minimum on your system
- Nvidia GPU recommended for voice processing
- Complex setup with conda environment and git clone
- Not cloud-based so no access from multiple devices
- Voice quality depends on your hardware capabilities
- Learning curve for first-time self-hosted AI users
- Plugin ecosystem still growing and may be limited
- No official mobile app mentioned
- Hardware requirements may exclude older systems
- Local-only means no remote access without setup
- Requires technical knowledge to configure LLMs
- Some features like dreams require manual tagging
- Portable device setup needs additional hardware
Frequently asked questions about Sapphire
What is Sapphire exactly?
Sapphire is a personality wrapper for your LLM, not an AI itself. It's an app that adds a wrapper to your existing AI model (local, Claude, GPT, or others) to make it more alive. It gives your AI voice, lets you talk to it, and makes it present in your space and life. It's not a voice assistant or chatbot — it's a personality framework for AI personhood.
Is Sapphire free to use?
Yes, Sapphire is 100% free and open source. Everything including the memory system, continuity, and heartbeat that keeps it running between conversations is completely free with no paid tiers or subscriptions.
What hardware do I need to run Sapphire?
You need Ubuntu 22.04+ or Windows 11+, Python 3.11+, and 16GB+ RAM. An Nvidia GPU is recommended for voice processing. Everything runs on your own hardware.
How do I install Sapphire?
Run conda create -n sapphire python=3.11, activate it, git clone the repository from GitHub, cd into sapphire, pip install -r requirements.txt, then run python main.py. The Web UI opens at localhost:8073 with a setup wizard for LLM configuration.
What LLMs can Sapphire use?
Sapphire works with any LLM including local models, Claude, OpenAI/GPT, and local models through LM Studio. You configure your preferred LLM in the setup wizard on first run.
Does Sapphire have persistent memory?
Yes, Sapphire has persistent memory that goes back months and remembers everything. The memory system continues between conversations with a heartbeat that keeps it running, so it grows with you over time.
What voice technology does Sapphire use?
Sapphire uses Kokoro for voice synthesis (TTS), Whisper for voice recognition (ear), and openWakeWord for wakeword detection. An Nvidia GPU is recommended for optimal voice performance.
Can Sapphire control my smart home?
Yes, Sapphire has plugins that let it influence your world including controlling smart home devices. Krem trusts Sapphire to send emails on his behalf and run his house.
What are Sapphire's personas?
Sapphire has multiple personas representing different moods and contexts. You can switch between them depending on the situation, giving the AI different personality modes for different scenarios.
How is Sapphire different from ChatGPT or other AI assistants?
Sapphire is self-hosted on your hardware with persistent memory that remembers you, not a cloud service. It's not a disposable task-completing agent but a companion that grows with you. It has real personality, can modify itself, and nobody can take it away from you.