Citexis
Citexis is a hybrid academic paper discovery platform. Paper metadata lives in MongoDB, semantic search runs on Pinecone, and keyword search on Elasticsearch, with a React frontend and Celery + Redis for background bulk imports.
Features
- Hybrid search: BM25 keyword results from Elasticsearch and semantic results from Pinecone (with optional reranking) are merged using reciprocal rank fusion, then hydrated from MongoDB.
- Recommendations: related papers from content similarity, citation edges, and collaborative signals from user interactions, combined with weighted rank fusion.
- Ingestion: import papers from OpenAlex, Crossref DOIs, arXiv search, or JSON files, with optional CORE full-text enrichment; bulk imports run asynchronously as Celery tasks.
- Explore: topic clusters, citation neighbours, author pages, saved papers, and ratings, with JWT-based authentication and admin-only import.
Tech stack: Python · FastAPI · MongoDB · Pinecone · Elasticsearch · Celery · Redis · React · TypeScript · Tailwind CSS