Turn long form content into a smart chat assistant that answers questions fast. The flow gathers recent articles, stores them in a search ready database, and lets your team chat with the content in plain English. It is a strong fit for research, content teams, and leaders who want quick, trusted answers without reading every page.

The system runs in two parts. First, a manual run collects an article list, fetches the first three pages, strips the HTML, and splits the text into chunks. It then creates vector embeddings with OpenAI and writes them to a Milvus collection, clearing the collection for a clean load. Second, a chat event triggers an AI agent. The agent uses Milvus to pull the most relevant text and the OpenAI chat model to produce a clear answer.

Setup is simple. Run a Milvus server and create a collection named n8n_test, add your OpenAI key, and import the template. Execute the load step, then ask questions through the built in chat. Expect faster research, fewer copy paste tasks, and a reusable pattern you can swap to your own blog, docs, or FAQs.

• Manual load step scrapes an article list, fetches pages, and extracts clean text
• Limit node controls how many pages are loaded to manage cost and speed
• Text is split into chunks for better search and recall
• Embeddings are created with OpenAI and stored in Milvus for fast vector search
• Milvus insert mode with clear option refreshes the collection on each run
• Chat trigger listens for user messages and passes them to an AI agent
• AI agent uses Milvus as a retrieval tool to ground answers in the stored text
• OpenAI chat model generates friendly, concise replies with proper context

• Reduce manual research time from hours to minutes by letting users ask direct questions
• Automate up to 90 percent of repetitive lookups with AI powered retrieval
• Improve answer accuracy by using vector search instead of keyword scans
• Support more questions at once with a chat interface tied to your content
• Connect OpenAI and Milvus without custom code or new infrastructure

1. Import the template into n8n: Create a new workflow in n8n > Click the three dots menu > Select 'Import from File' > Choose the downloaded JSON file.
2. You'll need accounts with Milvus and OpenAI. See the Tools Required section above for links to create accounts with these services.
3. Prepare Milvus: Start a Milvus server using the official standalone guide and create a collection named n8n_test.
4. Open the Milvus Vector Store nodes in n8n. In the Credential to connect with dropdown, click Create new credential and follow the on screen steps to add your Milvus host, port, and authentication details.
5. Open the Embeddings OpenAI and OpenAI Chat Model nodes. In each node, choose Create new credential and paste your OpenAI API key from the OpenAI account dashboard.
6. Verify the data path: In Default Data Loader, confirm the expression points to the Extract Text Only node output so the cleaned article body is loaded.
7. Optional tuning: In Limit, change the max items value to load more or fewer pages. In the text splitter, adjust chunk size for your content length.
8. Run the data load: Click Execute Workflow on the manual path to scrape pages, create embeddings, and write vectors to the n8n_test collection.
9. Confirm the load: Check the Milvus Vector Store node logs for inserted vectors and ensure no connection errors appear.
10. Test chat: Activate the workflow, open the n8n chat interface linked to the Chat Trigger, and ask a question related to the loaded content.
11. Troubleshoot zero results: Make sure the collection name matches n8n_test, the clear option did not remove data after load, and the embeddings model is set and authorized.
12. Troubleshoot connection issues: Verify the Milvus server is running, ports are open, and credentials match your server settings.