How b-s-f/yaku works
Project type and purpose
• Yaku is a full-stack, Kubernetes-native microservice platform and monorepo that automates software release evidence collection (“QG assessments”).
• It provides:
• A single-page UI (Vue + Vuetify) for human interaction (“qg-service-ui”).
• A REST API backend (NestJS “qg-api-service”)
• A CLI client (“yaku-cli” in TypeScript, “onyx” in Go) for pipelines or scripting
• A workflow engine (Argo Workflows) that spins up Docker-based “autopilot” fetcher/evaluator apps (Python, TS)
• Standard infrastructure: PostgreSQL, Keycloak, Minio/S3, Kubernetes/Helm
File structure
• Top-level directories show separation of concerns:
• chart/helm definitions for Kubernetes deployment of core-api, UI, labs
• qg-api-service/: NestJS monolith with modules/controllers/services, integration tests, database migrations, OSS dependencies
• yaku-ui/: two UI packages – frog-vue component library and qg-service-ui application
• onyx/: Go-based CLI tool for transformation, migration, execution pipelines
• yaku-cli/ and yaku-client-lib/: TS CLI and HTTP client library
• yaku-apps-python/ and yaku-apps-typescript/: dozens of containerized “autopilot” fetcher/evaluator utilities
• CI/CD workflows automate building Docker images for each component
README for additional insights
• The system uses Argo Workflows on Kubernetes to orchestrate jobs
• A PostgreSQL database named “yaku” stores core data
• Keycloak provides OIDC/SAML2 authentication