CodeWorker
An offline, Windows-only local LLM coding assistant that bundles llama.cpp, Python and Git into one folder, indexing a project with a local RAG search and code graph to answer questions and propose file edits without cloud access.
CodeWorker is a Windows local coding assistant built for offline, air-gapped or privacy-sensitive development, where source code cannot be uploaded to a cloud model. It bundles llama.cpp, a portable Python and Git distribution, GGUF models and a web UI in one working directory, so it needs no separate Ollama installation.
It detects the machine's RAM, CPU, GPU vendor, VRAM and Vulkan support to recommend a model and tuning parameters, and lets a user switch between several bundled GGUF models such as Gemma and Qwen variants. When a project is opened, it builds a local retrieval-augmented search index and a code graph of symbols, imports, calls and extends relationships, using them together with pinned files to answer questions or propose a reviewable edit plan. Writes, patches, deletions, renames and shell commands all go through a pending-action confirmation step, and a Git checkpoint is created before and after each applied change so results can be diffed and reverted.
It supports code, configuration, document, image, audio and video attachments, and a Codex plugin lets Codex query the same local code graph.