What is qwen-code?

qwen-code is a command-line coding assistant that runs entirely on your own machine, with no cloud API required. You interact with it through a terminal, either in an interactive back-and-forth session or by passing a one-off question as a command. It is described as a Claude Code-style tool, meaning it can reason through a task by repeatedly using tools to read files, run shell commands, search your codebase, and look things up before giving you a final answer. The AI model behind it is Qwen2.5-Coder, which runs locally via Ollama, a tool for running large language models on your own hardware. This means your code never leaves your machine. You can switch between different model sizes mid-session, change the working directory, or reset the conversation using slash commands inside the REPL. One of its notable features is RAG, which stands for retrieval-augmented generation. This lets you feed your own documents or code into a vector database called Qdrant, then have the assistant search that database when answering questions. It uses a separate embedding model called nomic-embed-text, also run through Ollama, to convert text into a searchable format. Qdrant can be run as a local server, a local file, or entirely in memory depending on your preference. The tool includes a permission system so that read-only operations like reading files or searching happen automatically, while anything that could change or delete files requires you to confirm with a yes or no prompt before it runs. The project is written in Python.