RAG Tool
Follow this workflow exactly when storing or retrieving document content.
Do not add verification steps unless the user explicitly asks.
Resolve Input
Always resolve and clean the input yourself before running cvstore or cvquery.
- Read plain text or markdown directly.
- For PDFs or other binary documents, use extracted text, not the binary file itself.
- Do not use scripts or programmatic pipelines to generate cleaned input text.
- Do not invent or infer content that is n
[Description truncada. Veja o README completo no GitHub.]