Phantasm

Phantasm is an iPhone AI chat app for people who want to use models running on their own machines. The phone provides the chat interface, while a backend you control handles inference.

• Local model chat: Connect to Ollama directly for plain chat or use an OpenAI-compatible backend.
• Streaming conversations: Read responses as they're generated, with Markdown formatting and images displayed in the conversation.
• Web search: An optional self-hosted orchestration server adds Brave search and page fetching so models can use information from the web.
• Image generation: The orchestration server connects to ComfyUI to generate pictures and display them in chat.
• Backend discovery: The app discovers models and detects supported capabilities, adapting to the services your server provides.

The native iOS client and optional server run on hardware you own. The server supports Docker Compose and can run alongside Ollama and ComfyUI on a home server, PC or Mac. Chat doesn't require an account or a cloud inference service, and the app doesn't collect telemetry. Web search contacts external services.

Your iPhone must be able to reach the backend over the network. Phantasm doesn't provide a managed remote-access relay.