Ollama_chat.nvim

Chat with local LLMs (like DeepSeek Coder) directly inside Neovim using an interactive markdown buffer and the Ollama CLI.

The chat lives inside a Markdown file, so code snippets and text formatting work naturally with tools like marksman or markdown-preview.

Perfect for in-editor AI coding assistance, quick explanations, or note-taking with models like deepseek-coder-v2, llama3, codellama, etc.

---

Features

• Interactive chat with Ollama in a Markdown buffer
• Auto-inserts ### User / ### Assistant blocks for clarity
• Live streaming responses, updated inline
• Cleans up ANSI escape sequences from output
• Minimal, no external dependencies

---

Installation

With lazy.nvim

{
  "Dheeraj-Murthy/Ollama_chat.nvim",
  config = function()
    require("ollama_chat").setup({
      model = "deepseek-coder-v2", -- optional, defaults to this
    })
  end
}

With packer.nvim

use {
  "Dheeraj-Murthy/Ollama_chat.nvim",
  config = function()
    require("ollama_chat").setup()
  end
}

---

Usage

1. Run :OllamaChat to open or jump to the chat buffer (OllamaChat.md).
2. Type your message under the ### User block.
3. Press <Enter> in normal mode to send it.
4. The model’s response will stream back under a new ### Assistant block.

---

Requirements

• Ollama installed (ollama run ... must work in your terminal)
• An Ollama-compatible model (e.g. deepseek-coder-v2, codellama, llama3)

Example to get started:

ollama run deepseek-coder-v2

---

Preview

▶️ Demo video: media/demo-fast.mp4

---

Roadmap

• Session history / persistence
• Model switching from inside Neovim
• Telescope integration to browse past sessions
• Richer Markdown formatting for roles

---

Author

M. S. Dheeraj Murthy
GitHub · LinkedIn

---

Contributing

Contributions are welcome — issues, PRs, and ideas!
If you build something cool on top (prompt templates, chaining commands, Telescope pickers, etc.), feel free to share it.