Skip to content

Tax Advisor

Tax Advisor is a Gemini-powered, tool-calling backend that helps individual investors make pre-trade tax decisions.

Code Video Cerebral Valley

Core trick

Gemini calls portfolio and realized-gains functions, then a second audit agent reviews the proposed tax-loss-harvesting plan.

Tags

financeagents

Description

Tax Advisor is a Gemini-powered, tool-calling backend that helps individual investors make pre-trade tax decisions. Instead of acting like a generic chatbot, it retrieves portfolio and realized-gain data through structured functions, then generates a quantified tax-loss-harvesting strategy. A second audit agent reviews risks such as timing, costs, and overconfidence, while a third agent creates an infographic summarizing recommendations for demo clarity. The service returns actionable priorities, projected tax impact, and transparent calculation limits. Built with Spring Boot, H2, and Cloud Run deployment, it demonstrates an end-to-end agent workflow: data retrieval, reasoning, auditing, and visual communication in one API for users.
jihyunmoon16/tax-advisor★ 00 forksJavaCreated Feb 28, 2026 · last push Feb 28, 2026

README excerpt

# tax-advisor

Spring Boot 기반의 MCP-style Agentic Tax Advisor 백엔드입니다.  
단순 질의응답 챗봇이 아니라, LLM이 함수 호출(`getUserPortfolio`, `getRealizedGains`)로 사용자 데이터를 직접 조회한 뒤 절세 전략을 수치로 제시합니다.
세금 계산은 해커톤 데모용 단순 모델(`과세표준 x 22%`)을 사용합니다.

## Project Description

### English

Tax Advisor is a Gemini-powered, tool-calling backend that helps individual investors make pre-trade tax decisions.  
Instead of acting like a generic chatbot, it retrieves portfolio and realized-gain data through structured functions, then generates a quantified tax-loss-harvesting strategy.  
A second audit agent reviews risks, and a third agent creates an infographic for clear communication.  
Built with Spring Boot and deployed on Cloud Run, it demonstrates an end-to-end agent workflow: data retrieval, reasoning, auditing, and visualization in one API.

### 한국어

Tax Advisor는 Gemini 기반 함수 호출(Function Calling) 백엔드로, 개인 투자자의 매도 전 절세 의사결정을 돕는 서비스입니다.  
단순 챗봇이 아니라 포트폴리오와 확정손익 데이터를 실제 툴로 조회한 뒤, 손실 실현(Tax-loss Harvesting) 중심 전략을 수치로 제시합니다.  
이어 감사 에이전트가 리스크를 점검하고, 디자이너 에이전트가 인포그래픽을 생성해 결과를 직관적으로 보여줍니다.  
Spring Boot + Cloud Run 기반으로, 데이터 조회부터 전략 생성·감사·시각화까지 하나의 API에서 처리하는 엔드투엔드 에이전트 워크플로우를 구현했습니다.

## 기술 스택

- Java 21 (Java 17+ 호환 구조)
- Spring Boot
- Spring Web / WebFlux(WebClient)
- Spring Data JPA
- H2
- Lombok
- Gemini API(Function Calling)

## 실행 전 준비

1. 루트에 `.env` 파일 생성 (`.env.example` 참고)

```bash
cp .env.example .env
```

`.env` 내용:

```bash
GEMINI_API_KEY=YOUR_GEMINI_API_KEY
GEMINI_MODEL=gemini-3-flash-preview
NANO_BA

Team