# Scholaris

> AI-powered bibliography and citation system for academic researchers. Converts PDFs, videos, audio, and images into searchable semantic documents (SPDF format). Runs 100% locally — no cloud processing, no data leaves your machine.

## Core Capabilities

- **Semantic Search**: Natural-language search across your entire research library. Cross-lingual (search in English, find results in Spanish). Cross-modal (one query searches PDFs, lecture recordings, and video simultaneously).
- **Auto-Citation**: Write your draft from memory, paste it in. Scholaris traces every claim to the exact page number in your library. Supports APA 7th, Chicago 17th, BibTeX.
- **SPDF Format**: Open document format built on SQLite. Stores full text, OCR, embeddings, images, transcripts, speaker segments, and metadata in a single portable file. Query with SQL. No vendor lock-in.
- **Privacy-First**: All AI models run locally on your hardware (NVIDIA or AMD GPU, or CPU-only). Documents never leave your machine. Designed for IRB-protected materials, unpublished manuscripts, and restricted archives.

## How It Works

1. Upload PDFs, video (MP4, MKV, WebM), audio (MP3, WAV, FLAC), images (PNG, JPG), or documents (DOCX, TXT, Markdown, EPUB).
2. Local AI processes everything: OCR (GLM-OCR), transcription (Parakeet/Whisper), speaker diarization (pyannote), semantic embeddings (Qwen3-VL).
3. Everything becomes a searchable SPDF file in your library.
4. Search by meaning across all media types. Get exact page numbers and timestamps.
5. Auto-cite your drafts with grounded, verified references.

## What Makes Scholaris Different

Elicit and Semantic Scholar search a cloud database of published abstracts to help you *discover* new research. Scholaris searches *your own library* — every page of every book, every minute of every recording — to help you *work with* sources you've already read. They say "this paper is relevant." Scholaris says "page 47, paragraph 3 directly supports your claim."

## Target Users

- Dissertation writers with hundreds of sources accumulated over years
- Humanities and social science researchers (books, not just papers; Chicago 17th citations)
- Seminar leaders with lecture recordings and reading lists
- Field researchers with IRB-protected interview recordings and archival materials
- Research groups sharing a collective library

## Pricing

- **Free**: Import and browse SPDF documents, organize into libraries, semantic search, export citations.
- **Pro** (€20/month, 14-day free trial): Unlimited conversions, AI-powered OCR, transcription, speaker identification, cross-modal search, auto-citation.

## Technical Stack

- Frontend: Next.js 14 (TypeScript, Tailwind CSS)
- Backend: FastAPI/Python (Uvicorn)
- AI Models (all local): GLM-OCR 0.9B, Qwen3-VL-Embedding 2B, Parakeet TDT 0.6B, pyannote speaker-diarization 3.1, Silero VAD, Qwen3 0.6B (query expansion), jina-reranker-v2 (cross-encoder reranking)
- Database: SQLite (SPDF files), ChromaDB (vector search)
- Auth: Clerk
- Payments: Stripe

## Links

- Website: https://scholaris.joseluissaorin.com
- Pricing: https://scholaris.joseluissaorin.com/pricing
- Privacy Policy: https://scholaris.joseluissaorin.com/privacy
- Terms of Service: https://scholaris.joseluissaorin.com/terms
- Full documentation for LLMs: https://scholaris.joseluissaorin.com/llms-full.txt