Skip to content

About

AI financial assistant that cites its sources — scam message checks, tax calculation, and housing subscription advice (LangGraph · GraphRAG · FastAPI · Next.js)

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

307 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Midas Touch

English | 한국어

CI

An AI financial assistant that shows its sources. It checks scam messages, calculates taxes, and answers housing subscription (청약) questions in one conversation, and every answer cites the law or public notice it is based on.

This is an information service, not investment or tax advice.

Features

Feature Description
Chat assistant (/chat) A LangGraph agent runs only the tools that match the question and streams the answer over SSE
Scam message check (/security) Rule-based verdict, response tips, and official report numbers. Includes a prompt-injection defense demo
Tax calculation Rates and deductions are computed by code, not the LLM, with National Tax Service articles attached as sources
Housing subscription (/cheongyak) Looks up notices from the public data API, calculates the 84-point score, and compares winning scores per notice
Savings simulator (/simulator) Estimates when a target amount is reached. Asset data is computed in the browser only
Knowledge graph (/graph) Tax law and subscription documents indexed in Neo4j and pgvector for GraphRAG retrieval

Architecture

Midas Touch architecture

Browser ── Vercel (Next.js) ── FastAPI + LangGraph ──┬─ PostgreSQL (pgvector)
                                (Oracle Cloud VM)    └─ Neo4j

Tech Stack

  • Frontend: Next.js 16, React 19, TypeScript, Tailwind CSS
  • Backend: Python 3.12, FastAPI, LangGraph, Alembic
  • Data: PostgreSQL 17 (pgvector), Neo4j 5
  • LLM: NVIDIA NIM (gemma-4-31b-it, bge-m3)
  • Infra: Vercel, Oracle Cloud VM (systemd, Caddy), GitHub Actions

Getting Started

uv sync
(cd frontend && npm install)
uv run alembic upgrade head
./dev.sh          # backend :8000, frontend :3000

See .env.example for environment variables. Tests and lint:

uv run pytest
uv run ruff check .

Project Structure

backend/     FastAPI routers and LangGraph agents
frontend/    Next.js web console
pipelines/   Data collection, document embedding, graph builder
shared/      DB clients, rate limiter
infra/       Caddy, systemd, auto-deploy scripts
tests/       pytest

Docs (Korean)

About

AI financial assistant that cites its sources — scam message checks, tax calculation, and housing subscription advice (LangGraph · GraphRAG · FastAPI · Next.js)

Topics

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages