The chatbot platform you can trust

How Does the RAG Astlai Work?

Retrieval-Augmented Generation (RAG) combines targeted document search with modern AI language capabilities. Unlike generic chatbots that answer "from the gut," Astlai first retrieves relevant facts from your databases—then generates a precise answer based on these sources. The result: reliable, verifiable answers instead of hallucinated "facts."

1

Document Preprocessing

We process your documents with intelligent chunking, create summaries, and extract metadata for optimal searchability

2

Intent Detection

We analyze the user query to understand their intention—this guides the pipeline toward more relevant results and determines whether retrieval is needed or if the model can answer directly

3

Multi-Stage Retrieval

We run retrieval multiple times in successive stages, each time getting closer to the required knowledge—starting broad with document summaries, then narrowing down to specific passages

4

Reranking

We rerank all retrieved results by actual relevance to the specific question, selecting the optimal information for the AI context

5

Response Generation & Citation

We generate a precise answer based on the retrieved sources and automatically insert citations that link every statement back to the original documents for full transparency