Build Your Own Qdrant Vector Store MCP Server

This n8n workflow demonstrates how to set up a Model Context Protocol (MCP) server that can build and manage a Qdrant vector store. It allows you to ingest data, split it into chunks, generate embeddings using OpenAI, and store them in a Qdrant instance.

What it does

This workflow acts as an MCP server, responding to requests to build a Qdrant vector store. Here's a step-by-step breakdown:

1. Triggers on MCP Server Request: The workflow is activated when an MCP Server Trigger receives a request.
2. Initializes Data: A "Sticky Note" node is present, likely for documentation or to hold initial configuration details for the vector store creation process.
3. Loads Data: A "Default Data Loader" node is used to load the input data, which will be processed and stored in the vector database.
4. Splits Text: The "Recursive Character Text Splitter" node breaks down the loaded text into smaller, manageable chunks. This is crucial for creating effective embeddings.
5. Generates Embeddings: The "Embeddings OpenAI" node uses the OpenAI API to convert the text chunks into numerical vector embeddings.
6. Stores in Qdrant: The "Qdrant Vector Store" node takes the generated embeddings and stores them in a Qdrant instance, creating or updating a vector collection.
7. Conditional Logic (Placeholder): An "If" node and a "Switch" node are present, suggesting that future enhancements might involve conditional processing or routing based on the input data or the result of the Qdrant operation. Currently, their specific conditions are not defined in the provided JSON.
8. Data Manipulation (Placeholder): "Edit Fields (Set)", "Filter", "Split Out", and "Aggregate" nodes are included, indicating potential data transformation or manipulation steps that could be integrated for more complex data handling before or after interacting with Qdrant.
9. External HTTP Request (Placeholder): An "HTTP Request" node is available, possibly for making external API calls related to the vector store or other services.
10. Code Execution (Placeholder): A "Code" node allows for custom JavaScript logic to be executed at various points in the workflow.
11. Manual Trigger (Placeholder): A "Manual Trigger" is included for testing or manually initiating the workflow.
12. Call n8n Workflow Tool (Placeholder): A "Call n8n Workflow Tool" node suggests the capability to call other n8n workflows as part of this process.

Prerequisites/Requirements

To use this workflow, you will need:

• n8n Instance: A running n8n instance to host the workflow.
• OpenAI API Key: For generating text embeddings. This will need to be configured in the "Embeddings OpenAI" node.
• Qdrant Instance: Access to a Qdrant vector database instance (either self-hosted or cloud-based). Connection details will be configured in the "Qdrant Vector Store" node.
• MCP Server Configuration: The MCP Server Trigger will need to be properly exposed and configured to receive requests from other systems or workflows.

Setup/Usage

1. Import the Workflow: Import the provided JSON into your n8n instance.
2. Configure Credentials:
   • OpenAI: Add your OpenAI API key as a credential in n8n and select it in the "Embeddings OpenAI" node.
   • Qdrant: Configure the connection details for your Qdrant instance (e.g., host, API key if applicable) in the "Qdrant Vector Store" node.
3. Customize Data Loading: Modify the "Default Data Loader" node to specify how your input data will be loaded (e.g., from a file, an API, a database).
4. Adjust Text Splitting: Review and adjust the parameters of the "Recursive Character Text Splitter" if needed, such as chunk size or overlap, to optimize for your specific data.
5. Activate the MCP Server Trigger: Ensure the "MCP Server Trigger" node is active and configured to listen for incoming requests.
6. Test the Workflow: You can use the "Manual Trigger" to test the workflow with sample data or send a request to the MCP Server Trigger from another n8n workflow or external application.