If you spend a meaningful amount of time working with MLIR in Neovim, you’ve probably noticed that syntax highlighting mostly exists for upstream dialects. Custom dialects are generally unsupported and treated almost like plaintext.
My journey of integrating Large Language Models (LLMs) into my daily development workflow continues to evolve. Recently, I’ve been focused on bridging the gap between Ollama, my preferred local LLM platform, and OpenCode. While the promise of a local, private coding agent is enticing, the reality of the initial setup was, frankly, underwhelming.
Lately I have been looking for more ways to integrate LLMs and agents into my workflow. Outside of the code I actually want to write, I am often faced with tasks I would categorize as laborious chores. This feels like a good fit for agents but I have also been thinking about privacy when using commercial offerings, so I have been exploring locally hosted small models.