RAG – FAISS (Facebook AI Similarity Search – In-Memory/Persistent)

Category Details
Key Features
  • Semantic search across architectural PDF drawings and documents.
  • Supports batch uploads and folder-level ingestion for large projects.
  • Displays match percentages and drawing previews for quick validation.
  • Search examples: “security office floor plan”, “ground floor layout”.
  • No paid or external APIs required by default — works with local/open-source components.
Tech Stack
Backend: Python 3.11+ (FastAPI, PyMuPDF, pytesseract, OpenCV, sentence-transformers, FAISS/ChromaDB).
Frontend: Node 18+ (Vite + React).
Testing: pytest (backend), Vitest (frontend).
CI/CD: GitHub Actions.
Use Case Designed for architects, civil engineers, and planners who need quick access to design elements inside large sets of architectural drawings and PDFs.
Extension: extract design-wise material reports for clients from the searched designs (bill-of-materials style summaries tied to drawing elements).