Abhi-lash19/VERA ? reverse-engineered prompt

Build me a terminal based knowledge assistant for engineering docs.

I want to drop in DOCX, PDF, and Markdown files, have it ingest them, and then ask natural language questions from the command line. It should search across the content intelligently, mix semantic and keyword matching, and give answers that are grounded in the files it found, with sources shown at the end. If a question is about a file or integration, it should be able to point me to the right place quickly. If I ask something like what is this flow, where is this config, or why is this failing, it should respond in a helpful, concise way using only the retrieved context.

Please build the full flow so I can run ingestion first, then query from the CLI. Keep all local indexes and metadata in my home folder, not inside the project. If you need to look up current docs online for any API details, go ahead.