llm-vscode
A VS Code extension for LLM-powered code completion, using llm-ls as its backend and configurable to call the Hugging Face Inference API, Ollama, Text Generation Inference, or any OpenAI-compatible endpoint.
llm-vscode is a VS Code extension for LLM-assisted development, using llm-ls as its backend. It supports Copilot-style ghost-text code completion, and a request can be sent to the Hugging Face Inference API, to Ollama, to Text Generation Inference, or to any OpenAI-compatible HTTP endpoint such as llama-cpp-python.
The prompt sent to the model is always sized to fit within the model's context window, using a tokenizer to count tokens. A code-attribution command checks whether generated code matches sequences of at least 50 characters in The Stack dataset using a Bloom filter, as a rapid first-pass check; a dedicated search tool over the full dataset is available for a complete second pass.
Related extensions exist for Neovim, Jupyter, and IntelliJ, and the project was previously named huggingface-vscode.