What is llamacore?

Llamacore is a desktop application that lets you run and chat with AI language models stored locally on your computer. It acts as a graphical front-end for a tool called llama.cpp, which is a widely-used program for running AI models without sending your data to an external service. The app is built with Electron, meaning it runs as a native window on Windows, macOS, and Linux. The core feature is a chat interface where you pick a local model file and start a conversation. Responses stream in word by word, and you can see live statistics like how fast the model is generating text. The chat window renders formatted text, code blocks, and math notation, and you can attach images if your chosen model supports vision. You can also edit any earlier message in a conversation to restart it from that point. Beyond chat, the app includes a model manager where you configure which local model files to use and start or stop the background server process each one needs. There is a conversion tool that takes a model from Hugging Face, a popular AI model repository, and converts it into the GGUF format that llama.cpp understands. A training monitor can display charts from a running TensorBoard session or from a training log file, refreshing every two seconds. The workflow editor is a visual node-graph tool where you connect steps together: feed text in, pass it through one or more AI calls, optionally run a shell command, branch on a keyword, and collect the result. Every shell command requires explicit confirmation before it executes, which the README flags as a security consideration. You are responsible for only running workflows you trust. All conversation history, model settings, and workflows are saved locally as JSON files on your own machine, not on any remote server. The interface supports both English and Simplified Chinese. The project is released under the MIT License.