calminer/docs/architecture/07_deployment_view.md

---
title: '07 — Deployment View'
description: 'Describe deployment topology, infrastructure components, and environments (dev/stage/prod).'
status: draft
---

<!-- markdownlint-disable-next-line MD025 -->

# 07 — Deployment View

## Deployment Topology

The CalMiner application is deployed using a multi-tier architecture consisting of the following layers:

1. **Client Layer**: This layer consists of web browsers that interact with the application through a user interface rendered by Jinja2 templates and enhanced with JavaScript (Chart.js for dashboards).
2. **Web Application Layer**: This layer hosts the FastAPI application, which handles API requests, business logic, and serves HTML templates. It communicates with the database layer for data persistence.
3. **Database Layer**: This layer consists of a PostgreSQL database that stores all application data, including scenarios, parameters, costs, consumption, production outputs, equipment, maintenance logs, and simulation results.

```mermaid
graph TD
    A[Client Layer] --> B[Web Application Layer]
    B --> C[Database Layer]
```

## Infrastructure Components

The infrastructure components for the application include:

- **Reverse Proxy (optional)**: An Nginx or Apache server can be used as a reverse proxy.
- **Containerization**: Docker images are generated via the repository `Dockerfile`, using a multi-stage build to keep the final runtime minimal.
- **CI/CD Pipeline**: Automated pipelines (Gitea Actions) run tests, build/push Docker images, and trigger deployments.
  - **Gitea Actions Workflows**: Located under `.gitea/workflows/`, these workflows handle testing, building, pushing, and deploying the application.
  - **Gitea Action Runners**: Self-hosted runners execute the CI/CD workflows.
  - **Testing and Continuous Integration**: Automated tests ensure code quality before deployment, also documented in [Testing & CI](07_deployment/07_01_testing_ci.md.md).
- **Docker Infrastructure**: Docker is used to containerize the application for consistent deployment across environments.
- **Portainer**: Production deployment environment for managing Docker containers.
  - **Web Server**: Hosts the FastAPI application and serves API endpoints.
  - **Database Server**: PostgreSQL database for persisting application data.
  - **Static File Server**: Serves static assets such as CSS, JavaScript, and image files.
- **Cloud Infrastructure (optional)**: The application can be deployed on cloud platforms.

```mermaid
graph TD
    G[Git Repository] --> C[CI/CD Pipeline]
    C --> GAW[Gitea Action Workflows]
    GAW --> GAR[Gitea Action Runners]
    GAR --> T[Testing]
    GAR --> CI[Continuous Integration]
    T --> G
    CI --> G

    W[Web Server] --> DB[Database Server]
    RP[Reverse Proxy] --> W
    I((Internet)) <--> RP
    PO[Containerization] --> W
    C[CI/CD Pipeline] --> PO
    W --> S[Static File Server]
    S --> RP
    PO --> DB
    PO --> S
```

## Environments

The application can be deployed in multiple environments to support development, testing, and production.

```mermaid
graph TD
    R[Repository] --> DEV[Development Environment]
    R[Repository] --> TEST[Testing Environment]
    R[Repository] --> PROD[Production Environment]

    DEV --> W_DEV[Web Server - Dev]
    DEV --> DB_DEV[Database Server - Dev]
    TEST --> W_TEST[Web Server - Test]
    TEST --> DB_TEST[Database Server - Test]
    PROD --> W_PROD[Web Server - Prod]
    PROD --> DB_PROD[Database Server - Prod]
```

### Development Environment

The development environment is set up for local development and testing. It includes:

- Local PostgreSQL instance (docker compose recommended, script available at `docker-compose.postgres.yml`)
- FastAPI server running in debug mode

`docker-compose.dev.yml` encapsulates this topology:

- `api` service mounts the repository for live reloads (`uvicorn --reload`) and depends on the database health check.
- `db` service uses the Debian-based `postgres:16` image with UTF-8 locale configuration and persists data in `pg_data_dev`.
- A shared `calminer_backend` bridge network keeps traffic contained; ports 8000/5432 are published for local tooling.

See [docs/quickstart.md](../quickstart.md#compose-driven-development-stack) for command examples and volume maintenance tips.

### Testing Environment

The testing environment is set up for automated testing and quality assurance. It includes:

- Staging PostgreSQL instance
- FastAPI server running in testing mode
- Automated test suite (e.g., pytest) for running unit and integration tests

`docker-compose.test.yml` provisions an ephemeral CI-like stack:

- `tests` service builds the application image, installs `requirements-test.txt`, runs the database setup script (dry-run + apply), then executes pytest.
- `api` service is available on port 8001 for manual verification against the test database.
- `postgres` service seeds a disposable Postgres 16 instance with health checks and named volumes (`pg_data_test`, `pip_cache_test`).

Typical commands mirror the CI workflow (`docker compose -f docker-compose.test.yml run --rm tests`); the [quickstart](../quickstart.md#compose-driven-test-stack) lists variations and teardown steps.

### Production Environment

The production environment is set up for serving live traffic and includes:

- Production PostgreSQL instance
- FastAPI server running in production mode
- Load balancer (Traefik) for distributing incoming requests
- Monitoring and logging tools for tracking application performance

#### Production docker compose topology

- `docker-compose.prod.yml` defines the runtime topology for operator-managed deployments.
- `api` service runs the FastAPI image with resource limits (`API_LIMIT_CPUS`, `API_LIMIT_MEMORY`) and a `/health` probe consumed by Traefik and the Compose health check.
- `traefik` service (enabled via the `reverse-proxy` profile) terminates TLS using the ACME resolver configured by `TRAEFIK_ACME_EMAIL` and routes `CALMINER_DOMAIN` traffic to the API.
- `postgres` service (enabled via the `local-db` profile) exists for edge deployments without managed PostgreSQL and persists data in the `pg_data_prod` volume while mounting `./backups` for operator snapshots.
- All services join the configurable `CALMINER_NETWORK` (defaults to `calminer_backend`) to keep traffic isolated from host networks.

Deployment workflow:

1. Copy `config/setup_production.env.example` to `config/setup_production.env` and populate domain, registry image tag, database credentials, and resource budgets.
2. Launch the stack with `docker compose --env-file config/setup_production.env -f docker-compose.prod.yml --profile reverse-proxy up -d` (append `--profile local-db` when hosting Postgres locally).
3. Run database migrations and seeding using `docker compose --env-file config/setup_production.env -f docker-compose.prod.yml run --rm api python scripts/setup_database.py --run-migrations --seed-data`.
4. Monitor container health via `docker compose -f docker-compose.prod.yml ps` or Traefik dashboards; the API health endpoint returns `{ "status": "ok" }` when ready.
5. Shut down with `docker compose -f docker-compose.prod.yml down` (volumes persist unless `-v` is supplied).

## Containerized Deployment Flow

The Docker-based deployment path aligns with the solution strategy documented in [Solution Strategy](04_solution_strategy.md) and the CI practices captured in [Testing & CI](07_deployment/07_01_testing_ci.md.md).

### Image Build

- The multi-stage `Dockerfile` installs dependencies in a builder layer (including system compilers and Python packages) and copies only the required runtime artifacts to the final image.
- Build arguments are minimal; database configuration is supplied at runtime via granular variables (`DATABASE_DRIVER`, `DATABASE_HOST`, `DATABASE_PORT`, `DATABASE_USER`, `DATABASE_PASSWORD`, `DATABASE_NAME`, optional `DATABASE_SCHEMA`). Secrets and configuration should be passed via environment variables or an orchestrator.
- The resulting image exposes port `8000` and starts `uvicorn main:app` (see main [README.md](../../README.md)).

### Runtime Environment

- For single-node deployments, run the container alongside PostgreSQL/Redis using Docker Compose or an equivalent orchestrator.
- A reverse proxy (Traefik) terminates TLS and forwards traffic to the container on port `8000`.
- Migrations must be applied prior to rolling out a new image; automation can hook into the deploy step to run `scripts/run_migrations.py`.

### CI/CD Integration

- Gitea Actions workflows reside under `.gitea/workflows/`.
- `test.yml` executes the pytest suite using cached pip dependencies.
- `build-and-push.yml` logs into the container registry, rebuilds the Docker image using GitHub Actions cache-backed layers, and pushes `latest` (and additional tags as required).
- `deploy.yml` connects to the target host via SSH, pulls the pushed tag, stops any existing container, and launches the new version.
- Required secrets: `REGISTRY_URL`, `REGISTRY_USERNAME`, `REGISTRY_PASSWORD`, `SSH_HOST`, `SSH_USERNAME`, `SSH_PRIVATE_KEY`.
- Extend these workflows when introducing staging/blue-green deployments; keep cross-links with [Testing & CI](07_deployment/07_01_testing_ci.md.md) up to date.

## Integrations and Future Work (deployment-related)

- **Persistence of results**: `/api/simulations/run` currently returns in-memory results; next iteration should persist to `simulation_result` and reference scenarios.
- **Deployment**: implement infrastructure-as-code (e.g., Terraform/Ansible) to provision the hosting environment and maintain parity across dev/stage/prod.