Fully autonomous multi-agent research system that assesses any country's AI capability across the full supply chain — from power grids and semiconductor fabs to AI models and talent pipelines.
Type a country name. 9 agents autonomously research, score, and compile a cited report. Zero human intervention.
A Supervisor Agent receives a country name and delegates research to 8 specialized Layer Agents, each targeting a different layer of the AI stack. Every agent searches live web data via the You.com Search API, scores its layer 0-100 relative to the global leader, and returns structured results with source citations. The Supervisor aggregates all scores into a weighted AI Power Score and classifies the country into a Power Tier.
| Layer | Weight | What It Measures |
|---|---|---|
| 1. Power & Electricity | 20% | Energy infrastructure for AI compute |
| 2. Chipset Manufacturers | 15% | Semiconductor design and fabrication |
| 3. Cloud & Data Centers | 15% | Compute infrastructure and capacity |
| 4. Model Developers | 10% | Foundation model research and training |
| 5. Platform & Data | 10% | Data accessibility and platforms |
| 6. Applications & Startups | 10% | AI ecosystem and commercialization |
| 7. Education & Talent | 10% | AI talent pipeline and research output |
| 8. Government Adoption | 10% | Policy, regulation, and public sector AI |
- Hegemon (80-100): Full-stack AI sovereignty — controls atoms and bits
- Strategic Specialist (50-79): World-class in specific layers, dependent on others
- Adopter (30-49): Growing ecosystem, follows leaders
- Consumer (0-29): Relies entirely on imported AI technology
├── NAAF-Research-Agents/
│ ├── agents/naaf_research/ # Google ADK agent system (primary)
│ │ ├── agent.py # Supervisor + sub-agent composition
│ │ ├── prompts/ # Layer-specific research prompts
│ │ ├── sub_agents/ # Layer agent factory
│ │ ├── tools/ # Search, scoring, persistence tools
│ │ ├── schemas.py # Pydantic output contracts
│ │ └── source_registry.py # Authoritative domains per layer
│ ├── reports/ # Generated research reports (JSON)
│ ├── webapp/ # React dashboard (Vite + Tailwind)
│ └── requirements.txt
├── backend/ # FastAPI backend + alternative agent impl
│ ├── agents/ # naaf_agent + naaf_multi agents
│ ├── api/app.py # REST API endpoints
│ ├── tools/ # You.com + Exa search integrations
│ └── framework/ # Layer definitions + scoring
├── web/ # Original web UI (React + Lovable)
├── docs/ # Documentation
│ ├── ai/ # AI engineering + prompt docs
│ ├── engineering/ # Technical architecture docs
│ ├── product/ # Product profile
│ └── internal/ # Session notes + handoffs
└── hackathon-submission/ # Devpost submission materials
cd NAAF-Research-Agents
# Create virtual environment
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
# Set API keys
export GOOGLE_API_KEY=your_gemini_api_key
export YDC_API_KEY=your_youcom_api_key
# Launch the ADK dev server
adk web agents/Open the ADK dev UI, select naaf_research, and type a country name to start an assessment.
cd backend
pip install -r requirements.txt
export YDC_API_KEY=your_youcom_api_key
export GOOGLE_API_KEY=your_gemini_api_key # optional
python run.py- Web UI: http://localhost:8000/ui
- API Docs: http://localhost:8000/docs
cd NAAF-Research-Agents/webapp
npm install
npm run dev| Component | Technology |
|---|---|
| Agent Orchestration | Google ADK + Gemini 3 Flash |
| Research Engine | You.com Search API (web + domain-restricted) |
| Data Layer | Supabase + structured JSON persistence |
| Backend API | FastAPI + Uvicorn |
| Frontend | React 18 / Vite / TypeScript / Tailwind CSS / Radix UI |
| Scoring | Weighted Relative Power Index with Pydantic schemas |
| Frontend Scaffolding | Lovable |
- You.com Search API — powers all agent research (web search + domain-restricted search)
- Google ADK + Gemini — agent orchestration, sub-agent delegation, session state
- Supabase — persistent data storage for the web dashboard
- Lovable — rapid frontend scaffolding and UI development
| Variable | Required | Description |
|---|---|---|
GOOGLE_API_KEY |
Yes | Google Gemini API key |
YDC_API_KEY |
Yes | You.com Search API key |
EXA_API_KEY |
No | Exa AI search key (optional) |
Pre-generated research reports are included in NAAF-Research-Agents/reports/:
- United States — Full 8-layer assessment with final report and source audit trail
- China — Partial assessment (layers 1-5)
- Singapore — Partial assessment (layers 1-4)
User Input (country name)
│
▼
┌─────────────────┐
│ Supervisor Agent │ ← Orchestrates research, aggregates scores
└────────┬────────┘
│ delegates to 8 sub-agents
▼
┌──────────────────────────────────────────┐
│ Layer 1 │ Layer 2 │ ... │ Layer 8 │
│ Power │ Chips │ │ Adoption│
└─────┬────┴─────┬────┴───────┴─────┬────┘
│ │ │
▼ ▼ ▼
┌──────────────────────────────────────────┐
│ You.com Search API │
│ (domain-restricted per layer) │
└──────────────────────────────────────────┘
│
▼
┌──────────────────────────────────────────┐
│ Scoring Engine │
│ Weighted RPI → Tier Classification │
└──────────────────────────────────────────┘
│
▼
┌──────────────────────────────────────────┐
│ Persistence Layer │
│ Layer JSONs → Final Report → Sources │
└──────────────────────────────────────────┘
MIT — Continual Learning Hackathon 2026