Couchbase + Dify: High-Power Vector Capabilities for AI Workflows

We’re excited to announce the new integration between Couchbase and Dify.ai, bringing Couchbase’s robust vector database capabilities into Dify’s streamlined LLMops ecosystem. Dify empowers teams with a no-code solution to build, manage, and deploy AI-driven workflows efficiently. Now, with Couchbase support, users can leverage Couchbase’s high-performance vector storage for rapid, accurate knowledge retrieval within their LLM-powered applications.

Dify.ai is an innovative platform that simplifies the process of building AI-native applications. It provides developers with a comprehensive suite of tools for prompt engineering, model fine-tuning, and application deployment, making it easier than ever to harness the power of large language models (LLMs) in practical applications.

In the following sections, we’ll explore some details of this integration and show how it can revolutionize building AI-native applications with Couchbase and Dify.ai.

Installing Dify with Couchbase

The following instructions are based on Dify’s self-hosted Docker Compose installation guide. Ensure the prerequisites on that guide are met.

1. 1. 1. Clone the Dify repository
         
         git clone https://github.com/langgenius/dify.git

         1	git clone https://github.com/langgenius/dify.git

      2. Navigate to the Docker directory and configure the environment
         
         cd dify/docker cp .env.example .env

         1 2	cd dify/docker cp .env.example .env

         
         IMPORTANT – in the duplicated docker/.env file, set:
         
         VECTOR_STORE=couchbase

         1	VECTOR_STORE=couchbase

      3. Launch Dify
         
         docker compose up -d

         1	docker compose up -d

      4. Access DifyAccess administrator initialization page to set up the admin account:
         
         # Local environment http://localhost/install # Server environment http://your_server_ip/install

         1 2 3 4 5

         # Local environment http://localhost/install   # Server environment http://your_server_ip/install

         
         Dify web interface address:
         
         # Local environment http://localhost # Server environment http://your_server_ip

         1 2 3 4 5

         # Local environment http://localhost   # Server environment http://your_server_ip

Instructions for upgrading Dify versions can be found here, and more details about Dify docker deployment can be found in the docker README.

Adding knowledge and generating embeddings

An embedding is a vector representation of text, enabling the “Knowledge” feature to understand and process the content.

1. 1. Access the Dify web interface
   2. Add model provider. This is required to generate embeddings:
      • Click on the avatar in the top right corner to expand the dropdown
      • Select Settings then Model Provider
      • Click the Setup button for the model provider you’d like to add. Avoid clicking “Add Model”, as this forces you to add each model one by one
      • Enter the required API keys, and click Save
      • Default system models can also be set from this window, but are not required
   3. Add knowledge:
      • Click on the Knowledge tab
      • Select Create Knowledge
      • Follow steps to upload a file
      • Select the defaults for Text Preprocessing and Cleaning
      • Click Save and Process
   4. Test Retrieval:
      • After the document has completed processing, click Go to document
      • Select Retrieval Testing from the sidebar
      • Experiment with various text to test the hitting effect of the Knowledge

Conclusion

The Couchbase and Dify integration makes it easier to build AI-native applications that harness high-performance vector retrieval for knowledge-intensive tasks. This powerful combination empowers developers to create sophisticated, responsive AI solutions with minimal setup, paving the way for more accessible and scalable AI applications.

Next steps

• • More information is available in the Dify documentation
  • See the guide to Creating Applications

Happy coding!