Resora is a cutting-edge, high-performance desktop ecosystem engineered to revolutionize the academic research lifecycle. Designed specifically for scholars, PhD candidates, and institutional research teams, it automates the most grueling aspects of literature reviews—from high-speed semantic discovery and systematic screening to neural synthesis and PRISMA-compliant reporting.
Resora transcends traditional reference management by providing a unified, AI-native workspace. It integrates advanced Natural Language Processing (NLP) with high-speed vector retrieval to solve the "Information Overload" problem in modern academia.
- 🧠 Semantic Discovery Engine: Leveraging FAISS (Facebook AI Similarity Search) and
all-MiniLM-L6-v2transformers to search 50,000+ papers based on conceptual meaning rather than rigid keywords. - 📝 Neural Synthesis (Summarization): Utilizing
BART-large-cnnarchitecture to condense dense academic abstracts into coherent, plain-English executive summaries. - ⚖️ LitRev-AI Systematic Screening: An advanced screening assistant using PubMedBERT-based classification to suggest Inclusion/Exclusion decisions, complete with a full decision audit trail.
- 📊 PRISMA 2020 Automation: Generate publication-standard PRISMA flow diagrams (PDF/PNG) dynamically based on your screening sessions.
- 🔍 Deep Paper Analysis: An automated extraction engine that identifies Research Methodologies, Contributions, Limitations, and Future Work from abstract text.
- 📅 Unified Research Workspace: A centralized hub for project-based organization, task management with deadlines, and multi-format citation exports (BibTeX, RIS, EndNote).
This repository contains the full 6th-semester NLP project submission. Below is the comprehensive directory map:
| Component | Path | Description |
|---|---|---|
| 🚀 Core Software | Code full Backend + Frontend/App |
The production-ready Python/PyQt5 application. |
| 📓 Model Training | Code Model Training & Other Things |
Jupyter notebooks for BERT fine-tuning and data preprocessing. |
| 📦 Artifacts | Code Output like model etc |
Fine-tuned model weights and system backup zips. |
| 💾 Data Assets | Data |
Metadata guides, data dictionaries, and reference dataset links. |
| 📄 Documentation | Proposal File & Report |
Formal project proposal, SRS, and 50+ page technical report. |
| 📸 Visual Media | Screenshots |
High-definition UI/UX walkthrough images. |
| 🎬 Demonstration | Video |
A 1080p full-length system demonstration video. |
| Component | Technology |
|---|---|
| Language | Python 3.11+ (Optimized for Windows/Linux/macOS) |
| GUI Framework | PyQt5 with custom QSS (Slate/Indigo SaaS aesthetics) |
| Vector Retrieval | FAISS (Facebook AI Similarity Search) with Inner Product indexing |
| NLP Core | Hugging Face Transformers, Sentence-Transformers, NLTK |
| Summarizer | facebook/bart-large-cnn (Conditional Generation) |
| Database | SQLite3 with WAL Journaling and Foreign Key constraints |
| Financials | Stripe API Integration (Subscription & Credit logic) |
| Visuals | Matplotlib (PRISMA flow generation) |
Windows users should use the provided setup_fix.bat to ensure CPU-only PyTorch is installed correctly, preventing common DLL errors.
- Navigate to App Folder:
cd "Code full Backend + Frontend/App" - Execute Fix Script:
./setup_fix.bat
- Run Application:
python main.py
cd "Code full Backend + Frontend/App"
pip install -r requirements.txt
python main.pyA robust project-management layer. Each project allows users to:
- Track paper reading status (Unread, Reading, Done).
- Assign task priorities (High, Medium, Low) with deadlines.
- Consolidate AI summaries and analysis results per project.
Designed for SLR (Systematic Literature Reviews):
- Heuristic + ML Scoring: Calculates a probability score for paper inclusion.
- Session Persistence: Save screening sessions to resume later.
- Export: Generate a full audit log in CSV and a PRISMA 2020 diagram.
An expert system specialized in academic queries:
- Explains complex statistical concepts (P-values, ANOVA, etc.).
- Suggests appropriate research methodologies based on study aims.
- Provides templates for writing hypotheses and research gaps.
A production-ready SaaS module:
- Stripe Checkout: Integrated payment flow for Pro/University plans.
- Promo Codes: Logic for
RESORA2026,PHD2024, etc. - Admin Dashboard: Manage user tiers, revenue tracking, and support tickets.
| User Type | Password | |
|---|---|---|
| System Admin | admin@trilit.ai |
admin123 |
| Guest User | N/A | Click "Continue as Guest" |
This repository uses Git LFS to manage high-dimensional vector data and model artifacts. Ensure you have Git LFS installed to pull the following critical files:
arxiv_faiss.index: 250MB+ vector index.arxiv_embeddings.npy: Pre-computed research embeddings.trilit_ai_vscode.zip: Full workspace backup.
- Multi-Agent RAG: Integration of local PDF parsing for full-paper Q&A.
- Cloud Sync: Optional PostgreSQL backend for team-wide collaboration.
- Zotero Integration: Direct sync with Zotero/Mendeley libraries.
Developed by Waqar Ali (waqi786) as a Project (6th Semester NLP). Institution: University Research Submission. License: MIT License.
Special thanks to Cornell University for the ArXiv dataset.