Deployment

Docker

The included docker-compose.yml runs the API with persistent index storage:

docker-compose up -d

This mounts artifacts/, docs/, and data/ into the container. The --seed flag auto-ingests sample documents on first boot.

Custom Dockerfile build

docker build -t askbase .
docker run -p 8080:8080 -v $(pwd)/artifacts:/app/artifacts askbase

Environment variables

Variable	Default	Description
RAG_ADMIN_TOKEN	admin-demo-token	Admin API token
RAG_USER_TOKEN	user-demo-token	User API token

Always override these in production.

Kubernetes

Manifests live in the k8s/ directory:

k8s/
  namespace.yaml        # Dedicated namespace
  backend-deployment.yaml   # API deployment + service
  frontend-deployment.yaml  # Frontend deployment + service
  ingress.yaml          # Ingress with TLS

Deploy

kubectl apply -f k8s/namespace.yaml
kubectl apply -f k8s/backend-deployment.yaml
kubectl apply -f k8s/frontend-deployment.yaml
kubectl apply -f k8s/ingress.yaml

Health probes

The deployment uses Kubernetes-native probes:

Probe	Endpoint	Purpose
Liveness	GET /health	Restart if the process is stuck
Readiness	GET /readyz	Only route traffic when the index is loaded

Scaling

Askbase is stateless at the API layer. The index file is read-only after ingestion, so you can scale replicas horizontally. Mount the index from a shared volume (PVC or S3-backed) for multi-replica setups.

# In backend-deployment.yaml
spec:
  replicas: 3
  template:
    spec:
      volumes:
        - name: index
          persistentVolumeClaim:
            claimName: askbase-index

Monitoring

The /metrics endpoint exposes Prometheus-compatible metrics. Add a ServiceMonitor or scrape config:

- job_name: askbase
  static_configs:
    - targets: ['askbase-api:8080']
  metrics_path: /metrics

Key metrics: request latency (p50/p95/p99), error rate, requests per second, index chunk count.

Production checklist

• Override default auth tokens via environment variables
• Mount index on persistent storage
• Configure ingress with TLS
• Set up Prometheus scraping for /metrics
• Run rag evaluate against your golden dataset after each re-index
• Set resource limits (256MB RAM is enough for most indexes under 100k chunks)