gsuuon/model.nvim

Supports both hosted (OpenAI, Google PaLM, Hugging Face) and local models (llama.cpp, Ollama) with provider-agnostic architecture, enabling seamless provider switching. Prompts are fully programmable in Lua with async multi-step execution, streaming responses directly into buffers with multiple edit modes (append/replace/insert), and persistent chat buffers with editable conversation history. Includes Treesitter syntax highlighting for chat files, context management via quickfix lists, and a growing vector store feature for semantic search capabilities.

396 stars. No commits in the last 6 months.