The author built a centralized, admin-controlled vector database service that allows multiple projects within an organization to embed and retrieve data from their own isolated vector stores. This approach eliminates the need for each project to maintain its own embedding logic and duplicate code.
- Each project receives its own named storage directory under a single base path, ensuring full isolation.
- The system uses a lightweight sentence-transformer model for embeddings and FAISS for fast similarity search.
- A Python backend handles chunking, upserting, and querying via simple endpoints.
- An admin layer provides full lifecycle control, including listing databases, deleting specific file vectors, and triggering steps via webhook.
This architecture turns setting up vector search into a quick integration task, reducing maintenance overhead by centralizing the codebase while keeping project data sandboxed.