How umami-dev/interview works
1. This repository is a mono-repo of three largely independent take-home assignments, not one unified product. Each top-level folder has its own runtime, dependencies, and deployment model, so the architecture is best drawn as three separate subsystems under a shared repo wrapper.
2. The root-level GitHub Actions workflow and CODEOWNERS file provide repository governance and CI for the repo as a whole. They do not appear to connect the three assignments at runtime; they mainly support validation, review ownership, and submission hygiene.
3. customer-data-pipeline is a Python batch/data-processing project. Its source lives under src/, with two main entrypoints: ingest.py and prepare_sync.py. The bundled input data is in src/data/input/, using JSONL files for customer profiles and membership records.
4. The pipeline’s data flow is file-in/file-out: read local JSONL inputs, transform and reconcile them, then produce a structured output that is compared against reference_output_structure.json. The presence of requirements.txt indicates a lightweight Python environment focused on offline processing rather than services or APIs.
5. dynamic-pricing is a Ruby on Rails API service. The main request path runs from config/routes.rb to app/controllers/api/v1/pricing_controller.rb, then into app/services/api/v1/pricing_service.rb, with shared service behavior in app/services/base_service.rb.
6. This Rails app has an explicit external dependency boundary in lib/rate_api_client.rb, which likely encapsulates calls to a third-party rate/pricing API. That makes the pricing workflow: incoming HTTP request → controller validation/dispatch → service orchestration → external rate client → response assembly. The included controller test suggests the public API is the primary contract.
7. The Rails project is container-friendly and operationally self-contained: Dockerfile, docker-compose.yml, config/puma.rb, and the standard Rails environment/config files define how it runs in development and production. The database and model folders exist, but the visible domain logic is concentrated in the API/service layer rather than persistence-heavy models.
8. terraform-parse is infrastructure-as-code, split across two provisioning layers. terraform/ contains Terraform definitions for cloud infrastructure, while helm/ contains Kubernetes manifests for frontend/backend deployments, services, and an HPA. This folder represents deployment and environment wiring rather than application logic.
9. The Helm chart indicates a deployed application boundary with separate frontend and backend workloads, each exposed by its own Service and scaled via the HPA. The Terraform layer likely provisions the supporting cloud resources needed by that Kubernetes deployment, but there is no shared runtime code with the other assignments.
10. A useful diagram should show three isolated vertical stacks beneath the repo: a Python batch pipeline over local files, a Rails API that may call an external rate service, and an IaC/deployment stack that defines cloud and Kubernetes resources. The only shared concern across them is repository-level CI and review metadata, not business data flow.
11. In short: customer-data-pipeline is an offline data transformation job, dynamic-pricing is a service boundary around pricing logic and an external rate API, and terraform-parse is infrastructure/deployment configuration. There is no evidence of direct coupling between these subsystems.
Main entry points: Chart.yaml Helm chart · ci.yml CI workflow · ingest.py Ingest · pricing_service.rb Pricing service
Components 20
Repo governance
Top-level CI and ownership metadata for the mono-repo.
- CI workflow · Repository-wide validation workflow.
.github/workflows/ci.yml - CODEOWNERS · Repository ownership and review routing.
.github/CODEOWNERS
Customer data pipeline
Offline Python batch processing assignment.
- Ingest · Reads local customer and membership inputs.
customer-data-pipeline/src/ingest.py - Prepare sync · Reconciles records and prepares structured output.
customer-data-pipeline/src/prepare_sync.py - Profiles · Customer profile JSONL input.
customer-data-pipeline/src/data/input/customer_profile.jsonl - Membership · Membership JSONL input.
customer-data-pipeline/src/data/input/membership.jsonl - Reference output · Expected output structure for comparison.
customer-data-pipeline/reference_output_structure.json
Dynamic pricing
Rails API assignment with external pricing integration.
- Routes · HTTP API route entrypoint.
dynamic-pricing/config/routes.rb - Pricing controller · Validates requests and dispatches pricing responses.
dynamic-pricing/app/controllers/api/v1/pricing_controller.rb - Pricing service · Coordinates pricing logic for the API.
dynamic-pricing/app/services/api/v1/pricing_service.rb - Base service · Common service behavior used by pricing orchestration.
dynamic-pricing/app/services/base_service.rb - Rate client · Encapsulates calls to the third-party rate API.
dynamic-pricing/lib/rate_api_client.rb - Rate API · Third-party pricing or rate data source.
Terraform parse
Infrastructure and deployment configuration assignment.
- Terraform · Cloud infrastructure provisioning layer.
terraform-parse/terraform/main.tf - Helm chart · Kubernetes deployment package for the application stack.
terraform-parse/helm/Chart.yaml - Backend deploy · Backend workload definition.
terraform-parse/helm/templates/backend-deployment.yaml - Backend service · Backend network exposure.
terraform-parse/helm/templates/backend-service.yaml - Frontend deploy · Frontend workload definition.
terraform-parse/helm/templates/frontend-deployment.yaml - Frontend service · Frontend network exposure.
terraform-parse/helm/templates/frontend-service.yaml - HPA · Horizontal pod autoscaling policy.
terraform-parse/helm/templates/hpa.yaml
Connections 19
- CI workflow → Ingest: checks · Repo CI validates the Python pipeline.
- CI workflow → Pricing controller: checks · Repo CI validates the Rails API.
- CI workflow → Terraform: checks · Repo CI validates IaC and deployment config.
- CODEOWNERS → CI workflow: governs · Ownership and review rules apply to repository workflows.
- Profiles → Ingest: input · Customer profiles are read by the ingest step.
- Membership → Ingest: input · Membership records are read by the ingest step.
- Ingest → Prepare sync: transforms · Initial file ingestion feeds reconciliation and sync preparation.
- Prepare sync → Reference output: matches · Pipeline output is compared to the expected structure.
- Routes → Pricing controller: dispatches · Rails routes map requests to the pricing controller.
- Pricing controller → Pricing service: delegates · Controller hands request handling to service logic.
- Pricing service → Base service: inherits · Pricing service shares common service behavior.
- Pricing service → Rate client: calls · Pricing orchestration invokes the external rate client.
- Rate client → Rate API: requests · Client boundary to the third-party rate API.
- Terraform → Helm chart: provisions · Terraform supports the Kubernetes deployment layer.
- Helm chart → Backend deploy: includes · Helm chart defines backend workload templates.
- Helm chart → Frontend deploy: includes · Helm chart defines frontend workload templates.
- Helm chart → Backend service: includes · Helm chart defines backend service templates.
- Helm chart → Frontend service: includes · Helm chart defines frontend service templates.
- Helm chart → HPA: includes · Helm chart defines autoscaling policy.