Summarsky

Summarsky is a book-processing and reading platform. It ingests source books, extracts structure and metadata, generates summaries and analysis artifacts, and exposes the results through both private dashboard routes and public reader routes.

Input and processing pipeline

The system supports EPUB, DOCX, ODT, RTF, Markdown, HTML, and plain-text inputs. Non-EPUB formats are normalized into EPUB-like processing inputs through a worker pipeline.

The documented processing flow is:

1. Validate quota and create an upload session.
2. Upload the file to storage.
3. Inspect duplicates and extract book-size estimates.
4. Reserve book-unit usage and enqueue work.
5. Worker downloads the source, normalizes file structure, extracts chapters and bibliographic metadata, and persists book records.
6. Worker performs chapter filtering and summarization.
7. Worker enqueues literary-analysis processing.
8. Reader and analysis routes expose the final artifacts.

Application architecture

The repository is split into:

• a Next.js web app,
• tRPC APIs,
• Prisma data models,
• a BullMQ worker,
• Redis-backed queueing,
• S3-compatible storage,
• Stripe billing,
• Auth.js authentication,
• OpenRouter/OpenAI-compatible model orchestration.

The project has a substantial operational layer:

• retained canonical source assets,
• book pipeline versioning,
• manual migration flows,
• archive import/recovery tools,
• admin diagnostics and audit history,
• duplicate matching and metadata backfill flows.

Reader model

The product exposes more than a single generated summary. Public and private reading surfaces include:

• book overview pages,
• chapter summary pages,
• chapter analysis pages,
• character detail pages,
• analysis pages for motifs, character arcs, and style/structure signals,
• dual-mode reading behavior that separates summaries from literary-analysis output.