Skip to content

DEPLOYMENT.md

Latest commit

 

History

History
99 lines (70 loc) · 1.69 KB

DEPLOYMENT.md

File metadata and controls

99 lines (70 loc) · 1.69 KB
Raw

Deployment

This project supports three deployment modes: SQLite local development, PostgreSQL with Docker Compose, and Kubernetes.

Local SQLite

pip install -r requirements.txt
python manage.py init-db
uvicorn ai_sql_analyst.main:app --reload

Docker Compose

docker compose up --build

This starts:

  • api: FastAPI application
  • postgres: PostgreSQL 16

The API is available at:

http://127.0.0.1:8000

Kubernetes

Render the manifests:

kubectl kustomize k8s/base

Apply:

kubectl apply -k k8s/base

The Kubernetes base includes:

  • Namespace
  • ConfigMap
  • Secret example
  • PostgreSQL StatefulSet and Service
  • API Deployment and Service
  • Liveness and readiness probes
  • HorizontalPodAutoscaler
  • Optional ingress example

For real environments, replace k8s/base/secret.example.yaml with a sealed secret, external secret, or managed secret provider.

Terraform

The Terraform example expects an existing Kubernetes cluster and local kubeconfig.

cd terraform/kubernetes
terraform init
terraform plan \
  -var="api_keys=replace-me" \
  -var="postgres_password=replace-me"

Terraform provisions:

  • Namespace
  • Secret
  • ConfigMap
  • PostgreSQL StatefulSet and Service
  • API Deployment and Service

Environment Variables

AI_SQL_ANALYST_DATABASE_BACKEND=postgres
AI_SQL_ANALYST_POSTGRES_DSN=postgresql://ai_sql:ai_sql_password@postgres:5432/ai_sql_analyst
AI_SQL_ANALYST_API_KEYS=dev-api-key
AI_SQL_ANALYST_BROWSER_API_KEY=dev-api-key
OPENAI_API_KEY=

CI

GitHub Actions runs:

  • Python tests
  • Text-to-SQL eval suite
  • Postgres-backed integration evals

Workflow:

.github/workflows/ai-sql-analyst-ci.yml