Building a $30,000 AI Knowledge Base Part 2: Upgrading Retrieval with Pinecone and Authentication Integration

Hey guys, just dropped Part 2 of building the $30,000 AI knowledge base. This one upgrades the filesystem retrieval from Part 1 to Pinecone so the chatbot searches by meaning instead of just keywords. Cover why Obsidian and Pinecone work together as different layers of the same pipeline, why just dumping all your notes into the AI stops working as the knowledge base grows, adding real authentication with Clerk, rotating the exposed API key, and shipping the whole thing to a live URL on Vercel. By the end you have a deployed AI knowledge base anyone can sign up to and query, built entirely with Claude Code. https://youtu.be/pm8UVKmBhvg