RAG Q&A Agent

RAG Q&A Agent

The retrieval-augmented answer pattern in four nodes: question → search → answer → output. Ground your LLM in your own documents and stop hallucinations at the source.

What this workflow does

1. Input receives the user question
2. RAG node (search mode) retrieves the top-K most relevant chunks from Qdrant
3. LLM node uses the retrieved context to answer accurately
4. Output returns the grounded answer

Use cases

• Internal knowledge base chatbot
• Policy and compliance Q&A
• Product documentation assistant

Setup

Configure the RAG node with your Qdrant collection (same one used in the RAG Document Ingest template). The LLM system instruction already references $RagSearch.context — connect your preferred model.

FAQ

Can I combine both RAG templates? Yes — run Ingest once per document, then re-use Q&A for every question.

Which embedding model should I use? Use the same model in both Ingest and Q&A to ensure vector compatibility.