System requirements
Memo's requirements depend on the models and features you choose. Running chat through an external provider has a different memory footprint from loading a large local GGUF model.
Desktop and server
Use the installer for Windows, Linux or macOS. Server-only installs also support Linux x86_64 and arm64, including Raspberry Pi. Check the platform package and engine availability before installing on a particular device.
Choosing a local model
The Model Store shows model size and a hardware-fit hint. Allow room for the model weights, context cache, the application and the embedding model. A model that barely fits on disk is not necessarily a good fit in memory. GPU support depends on the engine and hardware actually detected.
Memory and connected services
Persistent RAG needs an embedding model in addition to the chat model or provider. The setup wizard lets you skip embeddings, but explains that memory will not work without them. External providers, mail, messaging and remote tunnels require their respective network connections.
Mobile
The mobile app is a companion connected to a running Memo backend. The model and memory remain on that backend. An iOS IPA requires signing and sideloading.