Papan Ghosh

LocalLM turns an AI conversation into a practical workspace for getting things done. Developers and makers often have to copy information between a chatbot, an editor, a terminal, and external services. That breaks context and makes it harder to see what an AI actually changed. LocalLM brings those workflows into one desktop application. Powered by Meta's Muse Spark 1.3 for this submission, the assistant can work with selected workspace files and enabled tools, helping users understand a project, make changes, and inspect the results. An integrated editor, terminal, Git review, MCP connectors, and reusable skills connect the conversation to real work. The user controls the assistant's access through approval modes and permission rules. Tool decisions and results are recorded, file changes can be reviewed, and larger outputs are saved as inspectable artifacts. Built with Rust, Tauri 2, React, and TypeScript, LocalLM also supports compatible local GGUF models through llama.cpp. Users can choose local inference or a configured hosted provider. Muse Spark 1.3 runs through the hosted provider path; local inference is a separate option. The goal is an AI workspace that combines useful model capabilities with visible actions and user control.