AI Knowledge Base Guide
Train your AI with your own documents, websites, and Q&A pairs.
Overview
The Knowledge Base (RAG — Retrieval-Augmented Generation) lets you feed your own data to the AI. When a customer asks a question, the AI searches your knowledge base for relevant context and uses it to generate accurate, up-to-date answers.
Document Upload
Upload PDF, DOCX, TXT, JSON, or CSV files up to 50MB. Documents are automatically chunked (1000 characters, 200 overlap) and embedded using Cohere's multilingual model.
Website Scraping
Enter a URL and NevoChat will crawl and extract content from the website. Configure crawl depth (1-5 levels), delay between pages, or paste a custom sitemap.xml for precise URL targeting.
Q&A Pairs
Create question-answer pairs for precise, hand-crafted responses. These are embedded alongside document chunks for semantic search.
Semantic Search
Search your entire knowledge base with natural language queries. Enable AI reranking (Cohere rerank-multilingual-v3.0) for better result precision.
RAG Modes
Direct RAG: The knowledge base is always searched before every AI response. Best when answers should always reference your documents.
Tool-based RAG: The AI decides when to search the knowledge base. Best for conversational flows where not every message needs a document lookup.
Supported Content Formats
Using Knowledge Base in Flows
- 1.Upload your documents in Dashboard → Knowledge Base
- 2.In the Flow Builder, add a RAG Query node
- 3.Connect it before an AI Response node
- 4.The AI Response node will use retrieved chunks as context