gry67673905/-AI-

What it solves

This project provides a full-stack demonstration environment for a "One-Stop Government Service" AI assistant. It streamlines the process of accessing government services for citizens (individuals and enterprises) and staff, automating tasks such as service inquiries, eligibility pre-checks, material submissions, and application auditing while providing an AI-driven consultation interface.

How it works

The system is built as a modular monolith using FastAPI for the backend and a local WebView portal on Android for the frontend. It integrates several key components:

• AI Orchestration: Uses a combination of a local catalog, a Model Context Protocol (MCP) server for read-only government directory access, and RAG (Retrieval-Augmented Generation) for knowledge retrieval, powered by DeepSeek.
• Data Management: PostgreSQL serves as the authoritative source for services and applications, while Milvus handles vector embeddings for RAG, Redis manages session caching, and MinIO stores private materials.
• ** uma Android Portal**: A native Android wrapper that handles secure token storage (via Keystore) and integrates HMS (Huawei Mobile Services) for maps and navigation (Navi Kit).
• Mock Infrastructure: Includes a mock government API to simulate real-world government data and processes without requiring actual government platform access.

Who it’s for

• Developers and architects looking for a reference implementation of AI-integrated government service portals.
• Organizations wanting to prototype AI-driven administrative workflows (eligibility checks, document generation, and automated consultation).

Highlights

• Hybrid Knowledge Retrieval: Combines local catalogs, MCP servers, and RAG to provide grounded AI responses.
• Automated Document Generation: Asynchronously generates DOCX templates based on white-listed fields extracted by the LLM.
• Secure Architecture: Implements strict boundaries between public AI contexts and private user data, ensuring personal information does not enter LLM contexts.
• Full-Stack Demo: Includes everything from database migrations (Alembic) and container orchestration (Docker Compose) to a native Android APK build.