Document retrieval architecture

When a visitor asks a question that needs your knowledge base, Chatevo runs a retrieval pipeline before the LLM writes an answer.

Overview

User message
    → Smart query understanding (plan-gated)
    → Hybrid search (dense + keyword)
    → Top chunks + citations
    → LLM synthesis

Indexing

1. Extract text from PDF, DOCX, TXT, HTML, or crawled URL.
2. Chunk using rule-based and optional semantic chunking.
3. Embed chunks with an embedding model.
4. Store vectors in Qdrant with metadata.

Hybrid search

Retriever	What it finds
Dense (semantic)	Passages similar in meaning
Sparse (keyword)	Passages matching exact terms

Results merge via Reciprocal Rank Fusion (RRF).

Smart query understanding

Plan tiers: basic (Starter) · full (Standard+). See Smart query understanding.