Can I vibecode Tettra?

Build a deliberately narrow personal substitute for Tettra in an empty repository, focusing on an internal knowledge base and automated Q&A.
Use exactly this stack: Next.js 15, TypeScript, SQLite, and a basic semantic search implementation (e.g., using pgvector or a local vector SQLite extension).
Primary job: Build a single-user knowledge management workspace where users can write Markdown-based documentation, organize it into categories/projects, and use an AI-powered search bar to get instant answers based strictly on the documented knowledge.
Start from an empty folder and create the complete working project. Make the default mode single-user and private.
Store all documents and their vector embeddings locally in the self-hosted database.
Put every secret and external API credential (for embeddings/LLM) in .env and provide .env.example.
Implement a polished WYSIWYG or Markdown editor for creating and updating pages, supporting file uploads and internal linking.
Implement an "Ask the Wiki" feature: when a question is asked, retrieve the top relevant documentation chunks, inject them into an LLM prompt, and stream the generated answer with citations back to the source pages.
Include clear empty, loading, validation, success, and failure states. Add an explicit "Verification" state for pages (e.g., "Last verified 30 days ago") to keep content fresh.
Provide full workspace export as static Markdown files and a JSON metadata file.
Use accessible keyboard navigation, labels, focus states, and sensible contrast.
Deliberately exclude these paid-product advantages: real-time multiplayer editing, Slack/Teams bot integrations, enterprise SSO, and analytics on page views.
Write focused unit tests for the semantic search retrieval and the document tree data model. Add one end-to-end smoke test that proves a document can be created and queried.
Create a README with setup, permissions, architecture, data location, and limitations.
Finish by running the tests and listing the exact commands used.