macOS
codef requires an Apple Silicon Mac (M1 or newer) running macOS 13 or later. Download the DMG. Running local models with the bundled Ollama requires macOS 14 or later.
Windows
The EXE installer is built for Windows x64. Open Settings → System → About and check System type to find your architecture. To use local models, install Ollama separately and start it before selecting a model in codef.
Linux
The Linux x86-64 build is distributed as an AppImage. Allow the
file to run as a program in its properties. Launching it may
require FUSE 2. On Ubuntu 24.04 the corresponding package is
libfuse2t64.
To use local models, install and start Ollama. Saving a cloud provider key requires a system keyring such as GNOME Keyring or KWallet.
Memory and disk space
A cloud provider generates the replies for its models. A local model uses disk space for its files and memory while it runs. Leave extra disk space for the download and import when installing a GGUF file.
Memory use depends on the model, variant, and context size. If a model does not fit or runs slowly, try a smaller variant and reduce the context size in Settings → Models and API. How to choose a model size.