Modules Overview
Omelette is organized into eight core modules that form a research literature pipeline.
| # | Module | Description |
|---|---|---|
| 1 | Keywords | Three-level keyword hierarchy, LLM expansion, search formula generation |
| 2 | Literature Search | Federated search across Semantic Scholar, OpenAlex, arXiv, Crossref |
| 3 | Deduplication | DOI hard dedup, title similarity, LLM-verified dedup |
| 4 | Subscription | RSS feeds and API-based scheduled updates |
| 5 | PDF Crawler | Unpaywall, arXiv, direct URL fallback |
| 6 | OCR | pdfplumber + PaddleOCR for scanned PDFs |
| 7 | RAG Knowledge Base | ChromaDB vectors, hybrid retrieval, LLM answers with citations |
| 8 | Writing Assistant | Summarization, citations (GB/T 7714, APA, MLA), review outlines, gap analysis |
Pipeline Flow
Keywords → Search → Dedup → Subscription → Crawler → OCR → RAG → WritingEach module can be used independently or as part of the full pipeline. Projects organize literature; keywords drive search; results flow through dedup, crawl, OCR, and indexing before being queried for writing assistance.