What's the least annoying way to let AI search your own files?
Getting AI tools the right context from your own files still feels incredibly clunky. If an answer is buried inside local PDFs, markdown notes, or specs, manually digging through directories and copy-pasting paragraphs into the chat completely ruins the workflow.
I recently set up Linkly AI over MCP to handle this on my machine without forcing a full cloud migration.
Instead of dumping whole files into the context window, it indexes the folders and exposes them to the assistant as explicit tools. The model can search the map, inspect document outlines, and read precise text snippets on the fly only when needed. Files stay completely local, and the infrastructure runs flawlessly for my workspace.
I am running into a separate cross-platform headache, though. For those using local indexing or MCP tools, how do you handle real-time file-watching notifications (like inotify) over a network share like a local NAS via Samba or NFS? I want the indexer to catch edits across different machines instantly without relying on heavy network polling. Any specific mount configs you recommend?