Cognee Community Graph Adapter - Google Cloud Spanner (Spanner Graph)

This package provides a Google Cloud Spanner graph database adapter for Cognee using Spanner Graph (GQL). It allows using Cloud Spanner as the graph backend for knowledge graphs and GraphRAG pipelines when running in GCP.

Requirements

• Python >= 3.10, <= 3.13
• Google Cloud project with Spanner API enabled
• A Spanner instance (Enterprise or Enterprise Plus edition; Spanner Graph is not available on PostgreSQL dialect)
• A database created with the Cognee graph schema (see below)

Schema setup

Create a Spanner database with Google Standard SQL and run the following DDL to create the Cognee graph tables and property graph:

CREATE TABLE CogneeNode (
  id         STRING(MAX) NOT NULL,
  properties JSON,
) PRIMARY KEY (id);

CREATE TABLE CogneeEdge (
  source_id         STRING(MAX) NOT NULL,
  target_id         STRING(MAX) NOT NULL,
  edge_id           STRING(MAX) NOT NULL,
  relationship_type STRING(MAX) NOT NULL,
  properties        JSON,
) PRIMARY KEY (source_id, target_id, edge_id);

CREATE OR REPLACE PROPERTY GRAPH CogneeGraph
  NODE TABLES (CogneeNode)
  EDGE TABLES (
    CogneeEdge
      SOURCE KEY (source_id) REFERENCES CogneeNode (id)
      DESTINATION KEY (target_id) REFERENCES CogneeNode (id)
      LABEL Related
  );

You can create the database and apply this DDL via the Google Cloud Console, gcloud CLI, or the Spanner Admin API. See Set up and query Spanner Graph for details.

Installation

Create a virtual environment, then install the package (recommended):

cd packages/graph/spanner
python -m venv .venv
source .venv/bin/activate   # On Windows: .venv\Scripts\activate
pip install -e ".[dev]"   # Editable install with dev deps (pytest, pytest-asyncio)

Or install from PyPI (when published):

pip install cognee-community-graph-adapter-spanner

Or with uv:

uv venv && source .venv/bin/activate
uv pip install -e ".[dev]"

Usage

import asyncio
import cognee
from cognee.infrastructure.databases.graph import get_graph_engine
from cognee_community_graph_adapter_spanner import register


async def main():
    register()
    cognee.config.set_graph_database_provider("spanner")

    # Option A: single URL (project_id/instance_id/database_id)
    cognee.config.set_graph_db_config(
        {
            "graph_database_url": "my-project/my-instance/my-database",
        }
    )

    # Option B: separate parameters
    cognee.config.set_graph_db_config(
        {
            "project_id": "my-project",
            "instance_id": "my-instance",
            "database_id": "my-database",
        }
    )

    # Optional: custom credentials (defaults to Application Default Credentials)
    # cognee.config.set_graph_db_config({
    #     "project_id": "my-project",
    #     "instance_id": "my-instance",
    #     "database_id": "my-database",
    #     "credentials": your_credentials,
    # })

    await cognee.add(["Your content here."], "my_dataset")
    await cognee.cognify(["my_dataset"])
    results = await cognee.search(
        query_type=cognee.SearchType.GRAPH_COMPLETION,
        query_text="your query",
    )
    graph_engine = await get_graph_engine()
    nodes, edges = await graph_engine.get_graph_data()


if __name__ == "__main__":
    asyncio.run(main())

Configuration

Configure via set_graph_db_config() with one of:

Authentication uses Application Default Credentials unless credentials is provided.

Features

• Implements Cognee's GraphDBInterface using Spanner Graph (GQL) and SQL DML
• Async API; blocking Spanner calls run in a thread pool
• Node and edge properties stored as JSON
• Compatible with Cognee's add/cognify/search and graph visualization

Example

See the examples/example.py script for a full workflow (add data, cognify, search, graph data access). Ensure your Spanner database exists with the schema above and that ADC or credentials are set.

License

This project is licensed under the MIT License.