Skip to content

backend-connection-reference.md

Latest commit

 

History

History
120 lines (91 loc) · 4.67 KB

backend-connection-reference.md

File metadata and controls

120 lines (91 loc) · 4.67 KB
Raw

Backend Connection Reference — Intel® AI for Enterprise Agent Toolkit

All shared backends are deployed into their own namespaces and are reachable from any workload namespace in the cluster.

Redis (Session Memory & Caching)

Instance Kubernetes Service URL Scope
Redis Stack (deploy_redis=on) redis-stack-server.redis.svc.cluster.local:6379 (ClusterIP) redis://default:<password>@redis-stack-server.redis.svc.cluster.local:6379 All namespaces (in-cluster only)

Redis requires a password. The credentials and the full URL are stored in the redis-stack-server-credentials Secret:

# Full connection URL (preferred — do not hardcode the password)
kubectl get secret redis-stack-server-credentials -n redis \
  -o jsonpath='{.data.REDIS_URL}' | base64 -d

# Password only
kubectl get secret redis-stack-server-credentials -n redis \
  -o jsonpath='{.data.REDIS_PASSWORD}' | base64 -d

# Quick connectivity test
kubectl exec -n redis redis-stack-server-0 -- sh -c \
  'redis-cli -a "$REDIS_PASSWORD" --no-auth-warning ping'
# PONG

The default: username in the URL is required — Redis 6+ rejects the redis://:<password>@... form with WRONGPASS. The Service is ClusterIP and a NetworkPolicy is applied; Redis is not reachable from outside the cluster. Use kubectl port-forward for local access instead of changing the Service type.

Agent Sandbox (Sandboxed Execution)

Enabled with deploy_agent_sandbox=on. See agent-sandbox.md for full details.

Resource Value
Namespace agent-sandbox
Router service sandbox-router-svc.agent-sandbox.svc.cluster.local:8080
SDK pip install k8s-agent-sandbox
Default template python-sandbox-template (namespace: agent-sandbox) 
# Quick health check (requires port-forward or in-cluster)
kubectl port-forward -n agent-sandbox svc/sandbox-router-svc 8080:8080 &
curl -s http://localhost:8080/healthz
# {"status":"ok"}

OpenShell (Policy-Enforced Sandboxes)

Enabled with deploy_openshell=on (requires deploy_agent_sandbox=on). See openshell.md for full details.

Resource Value
Gateway namespace openshell-system
Gateway service openshell.openshell-system.svc.cluster.local:8080 (gRPC, mTLS)
Sandbox namespace openshell-sandboxes
Client mTLS bundle Secret openshell-client-tls in openshell-system
CLI OpenShell v0.0.116 release
GenAI provider genai-gateway (attach with openshell sandbox create --provider genai-gateway) 
# Admin access (ClusterIP only)
kubectl port-forward -n openshell-system svc/openshell 8080:8080 &

PostgreSQL + pgvector (Vector Store & Long-Term Memory)

Enabled with deploy_pgvector=on. See README.md for deployment details.

Resource Value
Namespace pgvector
Host pgvector.pgvector.svc.cluster.local
Port 5432
Database agentdb
User agentuser
Connection string postgresql://agentuser:<password>@pgvector.pgvector.svc.cluster.local:5432/agentdb
Credentials secret kubectl get secret pgvector-credentials -n pgvector -o jsonpath='{.data.DATABASE_URL}' | base64 -d

KubeRay (Distributed Computing)

Enabled with deploy_kuberay=on. Tuned via core/inventory/kuberay-config.yaml. See kuberay.md for the full guide.

Resource Value
Namespace ray-system (default; set via kuberay-config.yaml)
Ray Client API ray://ray-cluster-head-svc.ray-system.svc.cluster.local:10001
Ray Dashboard http://ray-cluster-head-svc.ray-system.svc.cluster.local:8265
Dashboard (local) kubectl port-forward svc/ray-cluster-head-svc 8265:8265 -n ray-system 
# Verify all Ray pods are Running
kubectl get pods -n ray-system

# Quick Python connectivity test (from inside the cluster or via port-forward)
python3 -c "
import ray
ray.init('ray://ray-cluster-head-svc.ray-system.svc.cluster.local:10001')
print(ray.cluster_resources())
"

Flowise (Agent Builder)

Enabled with deploy_flowise=on. See flowise.md for the full guide.

Resource Value
Namespace flowise
Flowise service flowise.flowise.svc.cluster.local:3000
Flowise admin UI https://flowise-<cluster_url> 
# Check pods
kubectl get pods -n flowise

# Retrieve the Flowise API key
grep '^flowise_api_key:' core/inventory/metadata/vault.yml | awk -F'"' '{print $2}'