OpenYak makes local file work visible

The pitch is strong because it targets the part most local-AI apps still miss: trust. If OpenYak can make file access, tool use, and progress legible, it moves from "cool demo" to something people might actually use daily.

• –Local-first file workflows are only credible when users can see what the agent touched, why it did it, and how to roll it back
• –The progress/todo panel matters more than it first sounds; it turns hidden agent work into an auditable process instead of a spinner
• –Ollama support is table stakes for this audience, but the real differentiator is making local documents feel native to the agent loop
• –The biggest test is whether it handles messy real-world Office files and folders without needing constant prompting or manual cleanup
• –Worth testing against setups like Ollama + Qwen, Llama, and smaller fast models to see whether the UX holds up across latency and quality tradeoffs