Logo Lanfrica

merezki-11/Naija-Rights

Domaine:

natural language processing

Type de record:

software
Créateur:
mer
Hôte:
Naija Rights is a production-ready RAG application bridging the gap between citizens and their constitution. Engineered with a FastAPI backend, local ChromaDB embeddings, and Gemini. Features a modern, mobile-responsive frontend that delivers cited legal answers in conversational Nigerian Pidgin. # Naija Rights A full-stack Retrieval-Augmented Generation (RAG) application designed to democratize access to the Nigerian Constitution. The system translates dense legal jargon into localized Nigerian Pidgin, providing a highly accessible conversational interface for the public to uncover their fundamental rights alongside official section citations. **Live Demo:** our-naija-rights.vercel.app **Backend API:** naija-rights.onrender.com --- ## Screenshots ### Welcome Screen ### Conversational Pidgin Interface ### Strict Section Citations ### Mobile View --- ## How It Works The system utilizes a custom RAG pipeline to ground every AI response strictly within the Nigerian Constitution: ```text User asks a legal question ↓ System embeds the query locally ↓ ┌────────────────────────────────────┐ │ │ │ ChromaDB Vector Store │ │ (Retrieves top 8 relevant chunks │ │ from the Nigerian Constitution) │ │ │ └──────────────┬─────────────────────┘ ↓ Prompt Engineer synthesizes chunks + user query (Enforces street-smart Pidgin persona & citation rules) ↓ Google Gemini 2.5 Flash LLM ↓ Response returned to user ``` **Example interactions:** - "Can police search my phone?" → System fetches and cites Chapter IV, Section 37 on privacy. - "If they arrest me on Friday, when should I go to court?" → System fetches and cites Section 35(4) on reasonable time. - "Can I do a peaceful protest?" → System fetches and cites Section 40 on peaceful assembly. --- ## Features - Localized AI Persona — strictly responds in natural, conversational Nigerian Pidgin - Official Citations — every response includes exact Chapter, Part, and Section references - Local Embeddings — uses all-MiniLM-L6-v2 via ChromaDB natively to bypass API rate limits - Intelligent Retrieval — fetches the top 8 most contextually relevant constitutional chunks per query - Responsive Glassmorphism UI — premium frosted-glass …