Obsidian and Pi Agent setup compared with Notion

The speaker explains why he returned from Notion to Obsidian after adopting terminal agents and local tools. His preference rests on direct access to Markdown files, which lets his existing tools work with the same notes. He still considers Notion and Confluence more polished for an all-in-one workspace, particularly for databases.

The walkthrough covers homelab documentation, infrastructure inventories and IP address records. Frontmatter supplies fields for table views, while templates define the format agents should follow. Pi Agent reads the vault to answer documentation questions and draft a VM onboarding plan. Reusable skills describe where information lives and which infrastructure workflows to use. The speaker says Pi supports local LLM models as well as existing subscriptions; local file storage alone does not establish that every demonstrated workflow stays offline.

MDBase query handles filtering and reporting on Markdown frontmatter, with examples of Markdown and CSV output. The speaker then demonstrates his open source Pi Agent plugin for Obsidian, which requires the Pi coding agent installed locally. Its four tool modes range from chat without CLI tools to shell execution. The current note becomes context automatically. At recording, the plugin lacks prompt support and visibility into background activity. The speaker says the code was entirely AI-generated and cautions that issues may remain.