abigailsylvie/african-ai-policy-kg

GraphRAG-powered knowledge graph that maps AI policies across Africa and suggests frameworks for countries without one # African AI Policy Knowledge Graph > *What if we could use existing African AI policies to help countries that don't have one yet?* That's exactly what this project does. --- ## The Problem Africa has 54 countries. Only a handful have a national AI policy — Kenya, Rwanda, Ghana, Zambia, and a few others. The rest? Nothing. No roadmap, no governance framework, no strategy. This is a problem because AI is already here, and countries without policies are navigating it blind. --- ## What This Project Does This project builds a **knowledge graph** from real African AI policy documents (PDFs), extracts key information from them, stores everything in a graph database, and then uses **GraphRAG** (Graph + AI) to suggest tailored AI policy frameworks for countries that don't have one. You ask: *"What should Chad's AI policy look like?"* The system looks at what Kenya, Rwanda, Ghana, and Zambia did, finds patterns, and generates a practical, country-specific suggestion. --- ## How It Works ``` PDF Policies (Kenya, Rwanda, Ghana, Zambia, African Union) ↓ Extract text with pdfplumber ↓ Extract entities with Groq AI (pillars, sectors, goals, risks, institutions) ↓ Store in Neo4j Knowledge Graph ↓ GraphRAG: query the graph + generate suggestions with LLM ↓ FastAPI REST API to access everything ``` --- ## Project Structure ``` african-ai-kg/ ├── data/ │ ├── policies/ ← Put your PDF files here │ ├── extracted/ ← Extracted text (auto-generated) │ └── entities/ ← Extracted entities (auto-generated) ├── src/ │ ├── extract.py ← Step 1: PDF → text │ ├── entity_extractor.py← Step 2: text → entities (Groq AI) │ ├── graph_loader.py ← Step 3: entities → Neo4j graph │ └── graphrag.py ← Step 4: graph + LLM = suggestions ├── api.py ← FastAPI REST API ├── main.py ← Run the full pipeline ├── requirements.txt └── .env.example ← Environment variables template ``` --- ## Getting St …