Enable Knowledge Graph

To enable Knowledge Graph in a self-hosted Prophecy deployment, follow these steps. 

Prerequisites

To enable Knowledge Graph in your Prophecy deployment, you need:

• Prophecy version 4.1.3.x or later.

• Access to your Kubernetes cluster with appropriate permissions.

• Images pulled for the services you are enabling:

  • Knowledge Graph: knowledgegraph, prophecy-kg-services

  • Neo4j: neo4j

Additionally, we recommend the following baseline resource allocation. Your resource requirements may vary.

Component	CPU (requests)	CPU (limits)	Memory (requests)	Memory (limits)
knowledgegraph	2	4	16000 Mi	16000 Mi
knowledgegraphservice	7	16	4000 Mi	4000 Mi
neo4j	1	4	4000 Mi	6000 Mi

Step 1: Enable services in configuration

Add the following services to all of the following locations:

• prophecycluster

• athena-pod-metadata-cm

enabledApps:
 - knowledgegraph, neo4j

To edit each resource, use the following commands:

kubectl edit prophecycluster -n <namespace>
kubectl edit cm athena-pod-metadata-cm -n <namespace>

Note: Every time you manually edit prophecycluster, you also need to edit athena-pod-metadata-cm to preserve these settings across Prophecy updates.

Step 2: Update cp-env-cm 

Edit cp-env-cm:

kubectl edit cm cp-env-cm -n <namespace>

Add the following environment variable:

COPILOT_KINESIS_ENABLED: "true"

Step 3: Update orchestrator-cm 

Edit orchestrator-cm:

kubectl edit cm orchestrator-cm -n <namespace>

Add the following for the Knowledge Graph:

KNOWLEDGEGRAPH_ENABLED: "true"
KNOWLEDGEGRAPH_PORT: "50055"
CRAWLER_SANDBOXCONFIG_CPU: "2"
CRAWLER_SANDBOXCONFIG_MAXTENANT: "50"
CRAWLER_SANDBOXCONFIG_MEMORY: 4Gi

Step 4: Restart Orchestrator Pod

To apply configuration changes, delete the orchestrator pod:

kubectl delete pod <orchestrator-pod-name> -n <namespace>

Alternatively, restart all pods (note: incurs ~20 minutes downtime):

kubectl rollout restart deploy -n <namespace>