Chat with Private Documents in n8n: Ollama + Qdrant

Chat with your private documents in n8n—powered by Ollama embeddings and Qdrant

This n8n workflow lets you upload confidential files, index them into Qdrant using Ollama embeddings, and then chat with an AI agent that answers questions by retrieving the most relevant passages from your stored documents.

What this workflow does

• Secure document upload: A basic-auth protected n8n form receives your uploaded documents.
• Prepare text for search: The workflow loads the file, attaches metadata (including the source filename and upload time), and splits text into chunks.
• Embed and index in Qdrant: Each chunk is embedded with Ollama using nomic-embed-text and inserted into a Qdrant collection (uses the private_documents collection name).
• Authenticated chat interface: A hosted chat UI requires n8n user login.
• RAG-style answering with citations: For every user message, an Ollama chat model (with conversation memory) searches Qdrant for relevant passages and returns a short answer based only on retrieved document content. It appends the source filenames, or reports it can’t find the answer in the documents.

Use cases

• Answer questions from internal policies, SOPs, or runbooks without exposing files to third-party SaaS chatbots.
• Enable support teams to query knowledge from uploaded documents while keeping responses grounded in the source text.
• Build a lightweight document Q&A feature for your organization using n8n, Ollama, and Qdrant.

Technical details

• n8n nodes: Form Trigger, LangChain Agent, LangChain Chat Trigger, Ollama Chat Model, Ollama Embeddings, plus supporting workflow steps.
• Ollama models: Example models include llama3.1:8b and nomic-embed-text:latest (editable to match your installation).
• Qdrant integration: Qdrant credential connects to your Qdrant instance and stores embedded passages for retrieval.