khReichel/obsidian-ai-agent — reverse-engineered prompt

Build me a small Python tool for my Obsidian vault that watches a folder of markdown notes, notices new or changed files, and automatically enriches them with AI. I want it to add an AI summary, AI tags, and AI keywords in YAML frontmatter, create semantic embeddings locally so the note content stays on my machine as much as possible, and detect related notes so it can write a Related Notes section plus matching AI metadata without overwriting my own tags or links.

It should remember what has already been processed so unchanged notes do not get reanalyzed, and it should handle API rate limits by waiting and retrying. I also want a simple terminal query mode where I can ask questions about my whole vault, either as a single command or in an interactive session.

Please make Docker the easiest way to run it, with a simple .env for the vault path and an LLM provider, ideally Groq by default but flexible enough for a few other providers too. Look up current docs online if you need to.