Sllm.nvim v0.2.0: On-the-fly Tools and a CLI-First Philosophy

I’m thrilled to announce the v0.2.0 release of mozanunal/sllm.nvim – a thin Neovim wrapper around Simon Willison’s amazing llm CLI. This release introduces powerful new features, most notably the ability to define LLM tools on-the-fly from your code, and further refines the plugin’s core philosophy.

The Neovim AI plugin space is indeed bustling! A fair question for any new entry is: where does it fit? This post aims to answer that by exploring the philosophy behind sllm.nvim and detailing what’s new in v0.2.0.

Why sllm.nvim? The Philosophy of a Focused Co-pilot

sllm.nvim isn’t designed to be a replacement for feature-rich plugins like CodeCompanion.nvim or Copilot.nvim, but a focused alternative built on a distinct philosophy. It was born from the desire to eliminate the constant context-switching to web UIs, by bringing the power of the llm CLI and its rich context management directly into the editor.

The philosophy is built on four key principles:

1. On-the-fly Function Tools: A Game-Changer This is perhaps the most significant differentiator. With <leader>sF, you can visually select a Python function in your buffer and register it instantly as a tool for the LLM to use in the current conversation. No pre-configuration is needed. This is incredibly powerful for interactive development:

   • Ad-hoc Data Processing: Have the LLM use your own function to parse a log file or reformat a data structure.
   • Live Codebase Interaction: Let the LLM use a function from your project to query a database or check an application’s state. This dynamic, ad-hoc workflow provides a level of integration that is unique to sllm.nvim.

2. Radical Simplicity: It’s a Wrapper, Not a Monolith sllm.nvim is a thin wrapper around the llm CLI (~500 lines of Lua). It delegates all heavy lifting (API calls, model management, conversation history, and tool integration) to Simon Willison’s robust, battle-tested, and community-maintained tool. This keeps sllm.nvim lightweight, transparent, and easy to maintain.

3. Instant Access to an Entire CLI Ecosystem By building on llm, this plugin instantly inherits its vast and growing plugin ecosystem.

   • Want to use OpenRouter’s 300+ models? Just run llm install llm-openrouter.
   • Need to feed a PDF manual or a GitHub repo into context? There are llm plugins for that. This extensibility comes “for free” and is managed at the llm level, not within the Neovim plugin.

4. Explicit Control: You Are the Co-pilot, Not the Passenger sllm.nvim believes in a co-pilot model where you are always in control. You explicitly provide context: the current file (<leader>sa), diagnostics (<leader>sd), the output of a command like git diff (<leader>sx), a URL, or a new function tool (<leader>sF). The plugin won’t guess your intentions, ensuring a predictable, reliable, and secure interaction every time.

What’s New in v0.2.0?

This release brings a host of improvements that double down on the core philosophy of deep llm integration and user control.

Headlining Features

• Register Tools (Functions) On-The-Fly (<leader>sF) (PR #41): As detailed above, you can now visually select a Python function and make it available to the LLM as a tool for the current session.
• LLM Tool Context Integration (llm -T) (PR #37): You can now browse your installed llm tools (from the CLI plugin ecosystem) and add them to the context for the LLM to use, all from within Neovim.

UI and UX Enhancements

• Configurable Window Type (PR #33): Choose between "vertical", "horizontal", or "float" for the LLM buffer to suit your workflow.
• UI Picker & Notifier Support (PR #35): Integrates with mini.nvim (pick/notify) and snacks.nvim (picker/notifier) for a more native and configurable UI experience.
• Better Window UI (PR #43): The LLM buffer now displays the model name and an indicator for running processes, providing more clarity at a glance.
• Visual Selection for Prompts (PR #51): You can now visually select text and use <leader>ss to send it along with your main prompt, making it easier to focus the LLM’s attention.

Core Improvements

• llm Default Model Support (PR #34): The plugin can now use the default model configured in your llm CLI setup, simplifying configuration.
• vim.ui.input Wrappers (PR #36): Better support for different input handlers for a more consistent experience.
• Lua Docs (PR #50): Added documentation for better maintainability and to help contributors understand the codebase.
• More Concise Preface & Agent Opinions (PR #55): The project’s philosophy documentation has been updated and clarified.

For a complete list of changes, check out the Full Changelog: v0.1.0 -> v0.2.0.

Get a Taste

Here’s a quick look at the on-the-fly function registration in action:

Try It Out!

If this CLI-first, co-pilot-centric philosophy resonates with you, I’d love for you to give sllm.nvim a try.

You can find the plugin, a full README, and installation instructions on GitHub: mozanunal/sllm.nvim

Your feedback, suggestions, and bug reports are invaluable. Let me know what you think! Thanks for reading.