RAG with Qdrant + LangChain

When to use RAG

Use Retrieval-Augmented Generation (RAG) when you need the AI to answer questions about specific content (manuals, an internal knowledge base, product documentation) without training a model from scratch. The LLM still generates the answer, but with context retrieved from a vector database.

Architecture

PDF / Markdown / Notion ──▶ embeddings ──▶ Qdrant
                                              │
       User question ──▶ embedding ──▶ search ──▶ context + LLM ──▶ answer

Stack

• Qdrant: vector database, runs in Docker
• LangChain (Python or Node): orchestrates chunks, embeddings and the prompt
• OpenAI ada-002 or self-hosted bge-m3: embeddings model
• Server: VPS Plus (8 GB RAM, 4 vCPU, 160 GB)

Next steps

We will soon publish the full tutorial with ready-to-use code. In the meantime:

• Install Ollama on a VPS (to run embeddings locally)
• n8n + EvolutionAPI + OpenAI (to connect RAG to WhatsApp)