How utsapoddar/leafmark works
Leafmark is a privacy-first, browser-based PDF/EPUB reading-guide app. The reader selects a book and optionally configures a model; the UI hands processing to the book processor, which extracts source-labeled text and produces snapshot, key ideas, chapter guide, and deep dive views. The default extractive path stays local; optional AI processing uses the semantic kernel and the reader’s selected provider. The graph reflects documented capabilities where implementation excerpts are incomplete; edges are limited to documented or sampled calls and data flows.
Main entry points: book-processor.ts Book processor · page.tsx Reader UI · model-providers.ts Provider client · semantic-kernel.ts Semantic kernel
Components 20
Reader workflow
- Reader UI
app/page.tsx - Privacy policy
app/privacy/page.tsx - Markdown export
app/page.tsx
Book processing
- Book processor
lib/book-processor.ts - Text extraction
lib/book-processor.ts - Local extractive engine
lib/book-processor.ts - Semantic kernel
lib/semantic-kernel.ts - Ledger checkpoints
lib/semantic-kernel.ts - Snapshot
lib/book-processor.ts - Key ideas
lib/book-processor.ts - Chapter guide
lib/book-processor.ts - Deep dive
lib/book-processor.ts - Source trail
lib/book-processor.ts
Model integration
- Provider setup
app/page.tsx - Provider client
lib/model-providers.ts
NVIDIA relay
- NVIDIA relay
relay/worker.ts
Other components
- Reader
- PDF or EPUB
- Model providers
- NVIDIA API
Connections 22
- Reader → Reader UI: uses
- Reader → PDF or EPUB: selects
- PDF or EPUB → Reader UI: provides
- Reader UI → Book processor: processes book
- Book processor → Text extraction: extracts text
- Text extraction → Book processor: returns segments
- Book processor → Local extractive engine: ranks locally
- Book processor → Semantic kernel: optional AI run
- Semantic kernel → Provider client: requests completions
- Provider client → Model providers: sends prompts
- Reader UI → Provider setup: configures model
- Provider setup → Provider client: discovers models
- Semantic kernel → Ledger checkpoints: reads and writes
- Book processor → Snapshot: produces
- Book processor → Key ideas: produces
- Book processor → Chapter guide: assembles
- Book processor → Deep dive: assembles
- Book processor → Source trail: retains references
- Reader UI → Markdown export: downloads guide
- Reader UI → Privacy policy: links policy
- Provider client → NVIDIA relay: forwards NVIDIA requests
- NVIDIA relay → NVIDIA API: proxies requests