Overview

Deploy Confident AI on your GCP infrastructure for complete control over your evaluation and observability platform.

Architecture

The GCP deployment uses Terraform to provision a production-ready infrastructure with the following components:

VPC & Networking

GKE/database/public/private-service-connect subnets with Cloud NAT, VPC firewall rules, and Cloud DNS private zone.

GKE Cluster

Managed Kubernetes with a system node pool and autoscaling worker node pool, Workload Identity, and VPC-native networking.

Cloud SQL for PostgreSQL

Managed database with Private Service Access in a VPC peering, regional HA, automated backups, and private DNS resolution.

NGINX Ingress + Cloud LB

NGINX Ingress Controller backed by a GCP TCP/UDP Network Load Balancer with cert-manager for TLS.

ArgoCD

GitOps tool for managing Kubernetes deployments.

Cloud Storage

GCS buckets with uniform access and Private Google Access for test cases, payloads, and ClickHouse backups.

The Kubernetes cluster runs the following services:

Service	Namespace	Description
confident-backend	confident-ai	Express.js API service handling core platform logic
confident-frontend	confident-ai	Next.js web application for the Confident AI dashboard
confident-evals	confident-ai	FastAPI service for running LLM evaluations
confident-otel	confident-ai	OpenTelemetry collector for trace ingestion
redis	confident-ai	In-memory cache for session and queue management
ArgoCD	argocd	GitOps continuous delivery
ClickHouse Operator	clickhouse-operator-system	Manages the analytics database
NGINX Ingress	ingress-nginx	Routes external traffic to services via Cloud LB
External Secrets	confident-ai	Syncs credentials from Google Secret Manager
cert-manager	cert-manager	Automates TLS certificate issuance and renewal