Detect polyphones, near-homophones, and ASR ambiguities. Two roles: (1) at runtime, extends the homophone table and the STT hallucination filter so the assistant doesn't get confused by "RAG vs LAG" or by streaming-platform watermark phrases; (2) in AI editors, gives the LLM a checklist for spotting ambiguities in user-generated knowledge bases.
Convert résumé / job descriptions / past interview notes into Markdown Q&A files in the format Interview Assistant's RAG expects (sections + Q/A pairs + optional Unicode tree blocks). Use when the user wants to populate `knowledge/` for the first time, ingest a new role, or refactor messy notes.