Skip to content

Repository files navigation

⚛ Plasma Intelligence

Free Stack Powered by Gemini License

Automated daily research intelligence for plasma physics. Every morning, Plasma Intelligence fetches new papers from arXiv and Semantic Scholar, analyzes them with Gemini 2.0 Flash, extracts figures, maintains living literature reviews across 18 subfields, and identifies open research problems — all for $0.00/day.


Features

Feature Description
📰 Daily Paper Digest Headlines + summaries for every new plasma physics paper
🔬 Deep Analysis Structured analysis: methodology, significance, key findings, open questions
📊 Figure Extraction Automatically extracts and describes figures from PDFs
📚 Literature Reviews Living, self-updating literature reviews for 18 subfields
Open Problems Extracts and clusters unsolved research questions across the field
🔍 Semantic Search Vector similarity search across all analyzed papers
📈 Importance Scoring AI-rated importance score (1-10) for every paper

Architecture

arXiv API + Semantic Scholar API
        ↓
FastAPI backend + APScheduler (daily 02:00 UTC)
        ↓
PyMuPDF (figures) + Gemini 2.0 Flash API (analysis)
        ↓
Supabase PostgreSQL + pgvector
        ↓
Next.js 14 frontend

Journals & Sources Covered

arXiv Categories: physics.plasm-ph, astro-ph.SR, astro-ph.HE, physics.space-ph, nucl-th

Via Semantic Scholar: Physics of Plasmas, Journal of Plasma Physics, Nuclear Fusion, Physical Review Letters, Physical Review E, Plasma Physics and Controlled Fusion, The Astrophysical Journal, Journal of Geophysical Research, and more.

18 Subfields Tracked

Magnetic Reconnection · Plasma Turbulence · Two-Stream Instability · Tokamak Physics · Stellarator Physics · Solar Wind Plasma · Magnetospheric Physics · MHD Waves · Kinetic Effects · Plasma Heating · Plasma Confinement · Plasma Transport · Collisionless Shocks · Accretion Disk Plasma · PIC Simulations · Drift Waves · Plasma Instabilities · Plasma Diagnostics

Prerequisites

Quick Start

See SETUP.md for complete step-by-step instructions.

# 1. Clone and configure
git clone https://github.com/Vivek-Shrivastav/Plasma-intelligence.git
cd Plasma-intelligence
cp .env.example .env
# Edit .env with your keys

# 2. Start
docker compose up -d

# 3. Test
cd backend && python ../scripts/test_pipeline.py

Daily Cost Estimate

Operation Volume Model Cost/day
Paper analysis ~80 papers Gemini 2.0 Flash $0.00
Figure description ~40 figs Gemini 2.0 Flash $0.00
Literature review patches ~80 Gemini 2.0 Flash $0.00
Open problems extraction ~80 Gemini 2.0 Flash $0.00
Total $0.00/day

API Endpoints

Endpoint Method Description
/api/papers GET List papers (filterable by date, subfield, importance)
/api/papers/{id} GET Full paper details + analysis
/api/subfields GET List all 18 subfields with stats
/api/subfields/{slug} GET Subfield details + recent papers
/api/literature/{slug} GET Literature review for a subfield
/api/open-problems GET Clustered open problems
/api/pipeline/trigger POST Manually trigger the daily pipeline
/health GET Health check

Tech Stack

  • Backend: Python 3.12, FastAPI, APScheduler, PyMuPDF
  • AI: Google Gemini 2.0 Flash (free tier)
  • Database: Supabase PostgreSQL + pgvector
  • Frontend: Next.js 14, React 18, TailwindCSS
  • Paper Sources: arXiv API, Semantic Scholar API

License

MIT

Releases

Packages

Contributors

Languages