Build and query knowledge graphs with Cognee
Cognee
Author: topoteretes
Version: 0.0.1
Type: tool
Description
Cognee is a memory plugin for Dify. It lets you ingest text data into datasets, build knowledge graphs with the Cognify engine, and search across them advanced retrieval techniques.
Setup
- Get your API key and base URL from your Cognee Cloud dashboard.
- Install the plugin in your Dify workspace.
- Configure the plugin with your Base URL (e.g. ) and API Key.
Tools
Add Data
Add text data to a Cognee dataset. Text can contain multiple items separated by newlines.
Parameters:
- Text Data (required) - Text content to add
- Dataset Name (optional) - Name of the target dataset. Either Dataset Name or Dataset ID must be provided.
- Dataset ID (optional) - UUID of an existing dataset. Either Dataset Name or Dataset ID must be provided.
- Node Set (optional) - Comma-separated node set names for graph organization
Cognify
Build a knowledge engine from one or more datasets. This processes the ingested data and may take several minutes depending on data volume.
Parameters:
- Datasets (optional) - Comma-separated list of dataset names. Either Datasets or Dataset IDs must be provided.
- Dataset IDs (optional) - Comma-separated list of dataset UUIDs. Either Datasets or Dataset IDs must be provided.
- Custom Prompt (optional) - Custom prompt for entity extraction and graph generation
- Ontology Key (optional) - Comma-separated ontology keys referencing previously uploaded ontology files
Search
Search the Cognee memory for relevant information.
Parameters:
-
Query (required) - Natural language search query
-
Datasets (optional) - Comma-separated list of dataset names to search
-
Dataset IDs (optional) - Comma-separated list of dataset UUIDs to search
-
Search Type (required, default: ) - Learn more: https://docs.cognee.ai/core-concepts/main-operations/search
-
System Prompt (optional) - System prompt for Completion-type searches
-
Top K (optional, default: 10) - Maximum number of results to return
-
Only Context (optional, default: false) - Return raw context instead of LLM completion
Delete Dataset
Delete an entire dataset and all its data permanently.
Parameters:
- Dataset ID (required) - UUID of the dataset to delete
Delete Data
Delete a specific data item from a dataset.
Parameters:
- Dataset ID (required) - UUID of the dataset
- Data ID (required) - UUID of the data item to delete
Usage in Dify Workflows
- Use Add Data to ingest text into a dataset.
- Use Cognify to build the knowledge engine from that dataset.
- Use Search before LLM calls to provide relevant context from memory.
- Use Delete Dataset or Delete Data to manage your datasets.