%20(1).jpeg){kind=link}
An AI-powered research synthesis platform that transforms information overload into actionable intelligence.
Orion harnesses Google's Gemini AI to conduct autonomous web research, analyze multi-modal documents, and generate publication-grade reports with unprecedented depth and nuance.
Traditional search engines return links. Orion returns understanding.
By combining large language models with advanced information retrieval, Orion doesn't just find sources—it reads them, synthesizes perspectives, identifies gaps, and produces coherent narratives that would take human researchers hours to compile.
- Adaptive Depth Control: Three synthesis modes (Précis, Synopsis, Treatise) dynamically adjust research scope and analytical rigor
- Multi-Modal Intelligence: Seamlessly processes text, PDFs, images, and structured data in a unified analytical framework
- Perspective-Aware Analysis: Automatically identifies competing viewpoints and presents balanced, evidence-based comparisons
- Citation-Integrated Output: Inline references and experimental APA formatting maintain academic rigor
- Real-Time Web Integration: Live SerpAPI connection ensures access to current information beyond training cutoffs
┌─────────────────────────────────────────────────────────┐
│ User Interface Layer │
│ (Streamlit + Custom CSS/HTML) │
└────────────────────┬────────────────────────────────────┘
│
┌────────────────────▼────────────────────────────────────┐
│ Application Logic Layer │
│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │
│ │ Web Research │ │ Document │ │ Conversation │ │
│ │ Pipeline │ │ Analysis │ │ Manager │ │
│ └──────────────┘ └──────────────┘ └──────────────┘ │
└────────────────────┬────────────────────────────────────┘
│
┌────────────────────▼────────────────────────────────────┐
│ AI Integration Layer │
│ ┌──────────────────────────────────────────────────┐ │
│ │ Google Gemini (Gemma 3 27B Instruct) │ │
│ │ • Dynamic prompt engineering │ │
│ │ • Context-aware generation (up to 8K tokens) │ │
│ │ • Multi-turn conversation management │ │
│ └──────────────────────────────────────────────────┘ │
└────────────────────┬────────────────────────────────────┘
│
┌────────────────────▼────────────────────────────────────┐
│ External Service Layer │
│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │
│ │ SerpAPI │ │ BeautifulSoup│ │ WeasyPrint │ │
│ │ (Web Search) │ │ (Scraping) │ │ (PDF Export) │ │
│ └──────────────┘ └──────────────┘ └──────────────┘ │
└─────────────────────────────────────────────────────────┘
Orion employs mode-specific prompt templates that dynamically incorporate user preferences:
- Précis Mode: Fast-track synthesis with executive summaries (100-500 words, ~4K tokens)
- Synopsis Mode: Balanced analytical reports with structured sections (1500-2500 words, ~8K tokens)
- Treatise Mode: Academic-grade research with abstracts and TOC (2000-4000 words, ~8K tokens)
Each mode integrates optional features (future projections, historical context, expert quotes, data visualization suggestions) through modular prompt injection—enabling unprecedented customization without combinatorial complexity.
Python 3.8+
pip (Python package manager)- Clone the repository
git clone https://github.com/yourusername/orion-research-engine.git
cd orion-research-engine- Install dependencies
pip install -r requirements.txt- Configure secrets
Create .streamlit/secrets.toml:
GEMINI_API_KEY = "your_gemini_api_key_here"
SERPAPI_KEY = "your_serpapi_key_here"
research_app_password = "your_secure_password"- Launch the application
streamlit run app.pyAccess at http://localhost:8501
streamlit>=1.28.0
google-generativeai>=0.3.0
requests>=2.31.0
beautifulsoup4>=4.12.0
markdown2>=2.4.0
weasyprint>=60.0
pypdf>=3.17.0
Pillow>=10.0.0
wordcloud>=1.9.0
matplotlib>=3.7.0
Query: "Ethical implications of large language models in medical diagnosis"
Output: Comprehensive analysis covering:
- Current deployment landscape
- Comparative accuracy studies (AI vs. human practitioners)
- Regulatory frameworks (FDA, EMA perspectives)
- Patient privacy considerations
- Future trajectory predictions with confidence intervals
Scenario: Upload 3 research papers on quantum computing
Capability:
- Cross-document synthesis
- Concept extraction and comparison
- Interactive Q&A with contextual memory
- Export conversation transcripts
| Feature | Description | Status |
|---|---|---|
| Perspective Analysis | Multi-viewpoint synthesis with proponent attribution | ✅ Active |
| Future Trajectories | Predictive insights based on trend analysis | ✅ Active |
| Data Viz Suggestions | AI-recommended charts/graphs for quantitative data | 🧪 Beta |
| Expert Quotations | Automated extraction of domain expert insights | 🧪 Beta |
| Historical Context | Temporal evolution tracking of research topics | 🧪 Beta |
- Password Protection: Three-attempt lockout mechanism
- Session State Management: Secure handling of user data
- No Data Persistence: Privacy-first architecture (no external storage)
Toggle between light/dark modes with custom CSS injection. Background images and color schemes are fully configurable via set_app_background().
- Inline Numbers:
[1],[2]superscript citations - Academic (APA): Experimental best-effort APA formatting
- None: Clean narrative output
| Metric | Précis | Synopsis | Treatise |
|---|---|---|---|
| Avg. Generation Time | 45-60s | 90-120s | 180-240s |
| Source Processing | 5-15 | 10-30 | 20-100 |
| Token Output | ~2K | ~5K | ~8K |
| Recommended Use | Quick overviews | Detailed analysis | Academic research |
Tested on standard queries with 15-100 web sources
- Academic Research: Literature reviews, gap analysis, hypothesis generation
- Corporate Intelligence: Market research, competitive analysis, trend forecasting
- Policy Analysis: Multi-stakeholder perspective synthesis, impact assessment
- Technical Due Diligence: Technology evaluation, risk analysis, vendor comparison
✅ Healthcare & Life Sciences
✅ AI/ML & Computer Science
✅ Finance & Economics
✅ Climate & Sustainability
✅ Regulatory & Legal Frameworks
We welcome contributions that enhance Orion's capabilities:
- Feature Requests: Open an issue with
[FEATURE]tag - Bug Reports: Detailed reproduction steps appreciated
- Pull Requests: Follow existing code style, include tests where applicable
- Multi-language support (beyond English)
- Graph-based source relationship visualization
- Automated fact-checking with confidence scores
- Export to LaTeX/Overleaf
- Integration with academic databases (PubMed, arXiv, IEEE)
This project is licensed under the MIT License - see LICENSE file for details.
Built with:
- Google Gemini API - Powering the intelligence layer
- SerpAPI - Enabling real-time web integration
- Streamlit - Rapid UI prototyping framework
- Open Source Community - BeautifulSoup, WeasyPrint, and countless other tools
Project Lead: [Your Name]
Email: your.email@domain.com
GitHub: @yourusername
Transforming information into insight, one query at a time.
⭐ Star this repo if Orion helped your research!