Sync n8n Docs to Qdrant with Ollama Embeddings (Workflow)

Keep your n8n documentation searchable—sync GitHub docs into Qdrant using Ollama embeddings

This n8n workflow automatically fetches the latest n8n-io/n8n-docs Markdown pages, converts them into vector embeddings with Ollama, and stores sectioned chunks in a Qdrant collection—so your documentation can be retrieved semantically.

What this workflow does

• Runs daily at 3:00 AM via an n8n Schedule Trigger.
• Detects changes by reading the latest GitHub commit SHA from n8n-io/n8n-docs and comparing it to a previously stored SHA (kept as a dedicated state point in Qdrant). If nothing changed, it can skip reprocessing.
• Prevents duplicates by deleting existing points from the n8n_docs Qdrant collection before re-ingesting.
• Fetches and filters docs by pulling the recursive GitHub file tree and selecting only docs/*.md files.
• Downloads raw Markdown content in batches, then splits each page into heading-based sections and enriches chunk metadata, including additional size-based sub-chunking.
• Embeds and indexes each chunk using Ollama with the nomic-embed-text model, then inserts the embedded documents into Qdrant under n8n_docs.
• Persists progress by writing the latest GitHub commit SHA back into Qdrant for the next run.

Use cases

• Build a documentation chatbot or RAG system that answers questions using n8n docs as the source of truth.
• Maintain an always-up-to-date semantic search index for n8n-io/n8n-docs without manual reprocessing.
• Enable SaaS operators to provide self-serve support by retrieving the most relevant doc sections from Qdrant.

Technical details

• GitHub integration: reads latest commit SHA and fetches Markdown files from n8n-io/n8n-docs.
• Qdrant: uses/creates a collection named n8n_docs configured for 768 dimensions and Cosine distance, with base URL defaulting to http://qdrant:6333.
• Ollama embeddings: generates vectors with nomic-embed-text via n8nn8n-nodes-langchainembeddings ollama.
• Vector store: inserts embeddings using n8nn8n-nodes-langchainvector store qdrant.
• n8n nodes used include code, sticky note, http request, schedule trigger, plus the LangChain Ollama/Qdrant nodes.