How Can I Map Qdrant Data Correctly for an Ollama Chat Agent in a RAG Workflow?

Below is some guidance to help resolve your issue.

Why Your Fields Are Empty

The Qdrant Vector Store node in n8n is designed to work with documents that follow LangChain’s document schema. By default, it expects the text content to be in a field called “pageContent” and any extra information in “metadata.” If your Qdrant documents use a different key (in your case, “text”), then when the node retrieves data it won’t find the expected fields and they’ll appear empty.

How to Resolve the Mapping Issue

The workflow design is correct for a typical RAG setup, but you’ll need to map your field names to what the Qdrant node expects. You can do this using a transformation step before passing data to the node. For example, add a Set or Function node in your workflow to convert the incoming field “text” to “pageContent” (and add any metadata as needed). Here’s an example using a Function node:

JAVASCRIPTCopied
// Assume the incoming item has a field "text"
items = items.map(item => {
  return {
    json: {
      pageContent: item.json.text, // mapping "text" to "pageContent"
      metadata: item.json.metadata || {} // ensure metadata exists if needed
    }
  };
});
return items;

Next Steps

1. Insert Transformation: Place a transformation node (Set or Function) before the Qdrant Vector Store node.
2. Update Field Names: Ensure that your documents have the proper keys:
   • Rename “text” to “pageContent”.
   • Provide any extra fields under “metadata”.
3. Test the Workflow: After mapping, test to see if the Qdrant node now retrieves the correct information.

Following these steps ensures that the data coming from Qdrant matches the expected input for subsequent nodes, allowing your RAG – AI Agent workflow to work as intended.

Let me know if you need any further details on how to set up the transformation.