RAG Electoral Law Assistant

A deployed, invitation-only application for source-grounded research across Mexican electoral law, regulations and approved reference material. The interface combines authenticated chat, retained conversation history, citations and role-based administration.

Legal research across multiple approved documents is slow and difficult to audit when answers cannot be traced back to the exact source material.

The application is private. This case study describes the engineering approach without exposing credentials, private data or protected source material.

I designed and built the TypeScript/Next.js application, ingestion and retrieval flow, authenticated product experience, usage controls and role-based administration.

The application ingests a manually approved corpus, retrieves relevant passages and generates grounded responses that retain reproducible citations. Protected storage and row-level access policies separate user and administrative capabilities.

• Authenticated Next.js progressive web application
• Server-backed conversation history and usage controls
• Approved-corpus ingestion and hybrid retrieval
• PostgreSQL and pgvector retrieval layer
• Protected document storage
• Supabase Row Level Security for user and administrative boundaries
• Citation-preserving answer generation

• Next.js
• Supabase Auth
• PostgreSQL
• pgvector
• Gemini
• Vercel

• Hybrid retrieval over an approved legal corpus
• Grounded answer generation
• Reproducible citations to underlying sources

• Use an approved source corpus rather than open-ended model knowledge.
• Preserve citations so every answer can be checked against the source material.
• Treat authentication, protected storage and role separation as product requirements.

• Authenticated chat and server-backed history
• Usage controls and role-based administration
• Document ingestion and hybrid retrieval
• Protected document storage and Row Level Security
• Grounded answers with retained citations
• Deployed private application

None listed.

Deployed private application

Turns paper-intensive legal research into fast, searchable and auditable access to approved source material.

Private application; architecture and implementation details are available for discussion.