vfarcic/youtube-automation

Purpose and Entry Points

YouTube Automation is a Go service and embedded React SPA for managing a YouTube video lifecycle from planning through publishing and post-publish work. The executable starts the HTTP server from cmd/youtube-automation/main.go; API contract is defined in openapi.yaml. The production server serves the built frontend embedded through internal/frontend/embed.go.

HTTP/API Boundary

internal/api/server.go composes routing, middleware, services, SPA fallback, and health handling. Resource handlers expose video CRUD, lifecycle phases, aspect metadata, AI actions, Drive operations, publishing, and social actions under authenticated /api/* paths. internal/api/middleware_auth.go applies optional Bearer-token authentication while health remains public. API handlers are the boundary between browser clients and domain/integration services.

Video Workflow and Persistence

Videos are represented as YAML-backed workflow records, with lifecycle phases and typed aspects driving progress and editing. internal/video/manager.go owns video-level management; internal/storage/yaml.go reads and writes persisted manifests; internal/workflow/types.go and internal/workflow/constants.go define workflow state. The service layer in internal/service/video_service.go and internal/data/video_service.go coordinates API-facing operations over this local data store, rooted at DATA_DIR.

Metadata-Driven UI and Forms

The aspect subsystem describes editable fields, mapping, setters, and completion calculations rather than hard-coding each video form. internal/aspect/service.go, internal/aspect/types.go, and internal/aspect/completion.go supply metadata and progress to API consumers. The React client renders that contract dynamically through web/src/pages/VideoDetail.tsx, web/src/components/forms/DynamicForm.tsx, and web/src/api/client.ts; the backend’s embedded build lives in internal/frontend/dist/index.html.

AI Content and Thumbnail Pipeline

AI adapters generate and analyze titles, descriptions, tags, tweets, shorts, AMA material, timing, and thumbnail prompts. internal/ai/provider.go establishes the provider boundary, while feature-specific operations include internal/ai/titles.go and internal/ai/analyze_timing.go. Thumbnail generation is a separate orchestration pipeline: internal/thumbnail/orchestrator.go coordinates prompts, generation providers, resolution, and image storage via internal/thumbnail/service.go and internal/thumbnail/image_store.go. AI credentials are externally configured (AI_KEY); Gemini/GPT image implementations are isolated in this layer.

Publishing and External Integrations

Publishing adapters turn video metadata and assets into external side effects. internal/publishing/youtube.go handles YouTube uploads, updates, captions, playlists, and analytics support; internal/publishing/hugo.go produces Hugo content/assets; internal/publishing/scheduler.go supports publish timing. Google OAuth and Drive are encapsulated by internal/auth/oauth.go and internal/gdrive/service.go. Social distribution is separated into platform adapters such as internal/platform/bluesky/bluesky.go, internal/platform/linkedin.go, and internal/platform/hackernews.go; Slack and email notifications use internal/slack/service.go and internal/notification/email.go.

Scheduled and Operational Runtime

AMA automation and notifications execute through an internal scheduler, centered on internal/scheduler/scheduler.go and internal/scheduler/ama_job.go. Runtime configuration comes from settings.yaml and internal/configuration/settings.go. Deployment supports a containerized Go binary (Dockerfile) and Kubernetes/Helm (helm/youtube-automation/templates/deployment.yaml), with the SPA built separately from the Vite React project under web/.