Knowledge Base

Overview

The Knowledge Base is TotalApp's Retrieval-Augmented Generation (RAG) system. You upload documents — PDFs, plain text, markdown — and TotalApp chunks them, converts each chunk to a vector embedding, and stores them in a searchable vector index. When an agent is configured to use a knowledge base, it queries this index at runtime and injects the most relevant chunks into its context window before generating a response.

Screen preview

A quick look at the screen. Use the toggle in the corner to switch between the dark and light themes.

Knowledge base: indexed sources

Planned Features

• Document upload — drag-and-drop interface for PDF, TXT, MD, and DOCX files. Batch upload up to 50 files at a time.
• Chunking strategy — choose between fixed-size chunks (e.g. 512 tokens) or semantic chunking that splits on paragraph or section boundaries.
• Collections — organise documents into named collections (e.g. "Product Manuals", "Legal Contracts"). Agents can query a specific collection or search across all of them.
• Query testing — enter a test query and see which chunks are retrieved before attaching the collection to an agent. Lets you validate that the retrieval is surfacing relevant content.
• Re-index — update a document in place without rebuilding the entire collection. Useful for frequently-updated source documents.
• Access control — mark collections as shared or private. Shared collections are available to all users' agents; private collections are only available to the creating user.