​
Configuration

While default options come directly with pip install cognee, dependencies for other databases should be installed from the optional-dependencies. For example, if you choose to run cognee with neo4j, pgvector, and postgres, this would look like pip install cognee[neo4j,postgres]. You can find all the optional dependencies here. After you make sure that all necessary dependencies are installed, you can configure your database variables.

​
🚀 Configure Vector and Graph Stores

You can configure the vector and graph stores using the environment variables in your .env file or programmatically. We use Pydantic Settings We have a global configuration object (cognee.config) and individual configurations on pipeline and data store levels Check available configuration options: 
from cognee.infrastructure.databases.vector import get_vectordb_config
from cognee.infrastructure.databases.graph.config import get_graph_config
from cognee.infrastructure.databases.relational import get_relational_config
from cognee.infrastructure.llm.config import get_llm_config
print(get_vectordb_config().to_dict())
print(get_graph_config().to_dict())
print(get_relational_config().to_dict())
print(get_llm_config().to_dict())
Setting the environment variables in your .env file, and Pydantic will pick them up: 
GRAPH_DATABASE_PROVIDER = 'lancedb'
Otherwise, you can set the configuration yourself: 
cognee.config.set_llm_provider('ollama')
Make sure to keep your API keys secure and never commit them to version control.

​
Settings for database engines

​
Relational Engines

The default SQLite does not need much (it can be omitted as well): 
DB_PROVIDER=sqlite
Postgres instance requires a bit more. Add the following to your .env file: 
DB_PROVIDER=postgres
DB_HOST=127.0.0.1
DB_PORT=5432
DB_USERNAME=cognee
DB_PASSWORD=cognee
DB_NAME=cognee_db

​
Vector Engines

​
LanceDB

LanceDB is the default vector database, it is a file-based vector store. To use it, add the following to your .env file (it can be omitted as well): 
VECTOR_DB_PROVIDER="lancedb"

​
PGVector

If you are already using Postgres, PGVector is a natural choice. 
VECTOR_DB_PROVIDER="pgvector"
If you’re using an AWS hosted Postgres you need to run CREATE EXTENSION vector; before first use. See official AWS docs

​
Qdrant

VECTOR_DB_PROVIDER="qdrant"
VECTOR_DB_URL=https://url-to-your-qdrant-cloud-instance.cloud.qdrant.io:6333
VECTOR_DB_KEY=your-qdrant-api-key

​
Weaviate

VECTOR_DB_PROVIDER="weaviate"
VECTOR_DB_URL=https://url-to-your-weaviate-cloud-instance.weaviate.cloud
VECTOR_DB_KEY=your-weaviate-api-key

​
Embedding Engines

Embedding engine is used to embed data before it is stored in a vector database. Cognee supports different embedding providers by implementing adapters for them. To configure it, add the following to your .env file: Example with Mistral: 
EMBEDDING_PROVIDER=mistral
EMBEDDING_MODEL=mistral/mistral-embed
EMBEDDING_API_KEY=mistral-api-key
EMBEDDING_MAX_TOKENS=8000
EMBEDDING_DIMENSIONS=1024
Example with Azure OpenAI Embeddings: 
EMBEDDING_PROVIDER=openai
EMBEDDING_MODEL=azure/text-embedding-3-large
EMBEDDING_ENDPOINT=https://azure-project.openai.azure.com/openai/deployments/text-embedding-3-large
EMBEDDING_API_KEY=azure-openai-api-key
EMBEDDING_API_VERSION=2024-12-01-preview
Example with Fastembed: Fastembed is used for the codegraph pipeline, because OpenAI models usually reach rate limit when ingesting the codebase. 
EMBEDDING_PROVIDER=fastembed
EMBEDDING_MODEL=sentence-transformers/all-MiniLM-L6-v2
EMBEDDING_DIMENSIONS=384
EMBEDDING_MAX_TOKENS=256

​
Graph Engines

​
Kuzu

KuzuDB is the default database we use. For Kuzu graph database, add the following to your .env file: 
GRAPH_DATABASE_PROVIDER="kuzu"
Its persistance is slightly different. It will write files to the .cognee_system/databases/cognee_graph.pkl directory.

​
Neo4j

Neo4J needs a bit more setup, a default for a locally hosted Neo4j instance would look like this: 
GRAPH_DATABASE_PROVIDER="neo4j"
GRAPH_DATABASE_URL=bolt://localhost:7687
GRAPH_DATABASE_USERNAME=neo4j
GRAPH_DATABASE_PASSWORD=pleaseletmein
If you installed your Neo4J manually, don’t forget to install apoc and graph-data-science plugins.
neo4j-local For the cloud instance, add the following to your .env file: 
GRAPH_DATABASE_PROVIDER=neo4j
GRAPH_DATABASE_USERNAME=neo4j
GRAPH_DATABASE_URL=neo4j+ssc://12345678.databases.neo4j.io
GRAPH_DATABASE_PASSWORD=neo4j-api-key

​
NetworkX

NetworkX is a file-based graph storage. For NetworkX graph database, add the following to your .env file: 
GRAPH_DATABASE_PROVIDER="networkx"
It will persist the data in .cognee_system/databases/cognee_graph.pkl JSON file.

​
Hybrid Engines

​
FalkorDB (Graph + Vector)

FalkorDB can be used as both the vector store and the graph database. Add the following to your .env file: 
VECTOR_DB_PROVIDER=falkordb
VECTOR_DB_URL=localhost
VECTOR_DB_PORT=6379
GRAPH_DATABASE_PROVIDER=falkordb
GRAPH_DATABASE_URL=localhost
GRAPH_DATABASE_PORT=6379

​
Amazon Neptune Analytics (Graph + Vector)

Amazon Neptune Analytics can be used as a hybrid backend providing both graph storage/traversals and vector search. Add the following to your .env file: 
GRAPH_DATABASE_PROVIDER="neptune_analytics"
GRAPH_DATABASE_URL=neptune-graph://g-your-graph
VECTOR_DB_PROVIDER="neptune_analytics"
VECTOR_DB_URL=neptune-graph://g-your-graph

AWS_DEFAULT_REGION=us-east-1
AWS_ACCESS_KEY_ID=your-access-key
AWS_SECRET_ACCESS_KEY=your-secret-key
AWS_SESSION_TOKEN=your-session-token
Ensure that the Neptune Analytics vector dimension matches your embedding model (for example, 1536 for OpenAI text-embedding-3-small).