Cognee + Continue Integration

1. Download the Cognee environment file  from the Cognee repository.

2. Fill in the necessary details in the environment file.

   NOTE: Remove all quotations marks ( ” ” ) in the file! DB_PROVIDER=“sqlite” will not work, it must be for example DB_PROVIDER=sqlite

   Make sure to also add the following embedding settings to the file:

   EMBEDDING_PROVIDER=fastembed
   EMBEDDING_MODEL=sentence-transformers/all-MiniLM-L6-v2
   EMBEDDING_DIMENSIONS=384
   EMBEDDING_MAX_TOKENS=256

3. Save the file in a location of your choice.

4. Note down the path to this file as you will need it for the next steps.

1. Open your preferred IDE

2. Install the Continue extension following the official documentation 

3. Create and log in to your Continue Hub account

4. Set necessary configurations and secrets in Continue for Continue to work with cognee. Cognee specific environment secrets explained:

   • COGNEE_ENV_FILE_PATH: Points to the location of your Cognee environment file, which contains configuration settings for the Cognee MCP server. This file typically includes API keys, model preferences, and other operational parameters. This file must be filled properly for Cognee to work.
   • COGNEE_SHARED_DIR_PATH: This parameter specifies a local directory for data sharing between Cognee and Continue. It plays a critical role in exchanging knowledge graph details, caching results, and preserving session state. To enable Cognee to codify the repository you’re working on, the repo must be at this shared path location.

5. Verify the installation by opening the Continue panel in your IDE

Note: Continue will need to be configured to work with AI providers after installation

Continue offers two integration methods for working with cognee:

1. Pre-built Assistant Option:

   • Use the ready-made cognee assistant
   • Access standard knowledge graph capabilities

2. Custom Assistant Option:

   • Build your own using the Cognee Context Provider block
   • Customize the integration to your specific needs
   • Requires additional configuration

Choose your preferred method and proceed with:

1. Open VSCode and use SHIFT-COMMAND-P (Mac)
2. Select Continue: Open in new window
3. In the top right corner, choose either:
   • Pre-built cognee assistant
   • Your custom assistant with Cognee Context Provider

Restart your IDE to ensure the updated Continue configuration is applied. You should see Continue recognizing the cognee integration on startup.

Prompt Continue with exactly the following:

Codify the /data repo

If everything is properly connected it will start the codify process on your repo provided with COGNEE_SHARED_DIR_PATH. The /data directory in the prompt is where your repo has been stored in the Cognee docker container.

To ask questions about your codebase you can use the following kind of prompt:

CODE search for function names

1. Open your project and start a new Continue session to generate a knowledge graph:

   This prompt will let Cognee know to use the CODE search type to ask about function names. Any other question can be asked about your code as well. For example:

   CODE search for class names

   The prompt above will return all classes in your code base

2. Get more accurate and contextually aware suggestions

   

If you want to make custom changes to Cognee and the Cognee MCP Server, you can deploy it locally. This is an optional step if you are satisfied with the default Cognee configuration. Install Cognee MCP Server