Skip to content

Bonk1211/Finlytics

BranchesTags

Repository files navigation

BorneoHack FastAPI Next.js Gemini LangGraph

🌏 BorneoHack — ASEAN Cross-Border Trade & Fintech Platform

An AI-powered platform empowering ASEAN MSMEs (Micro, Small, and Medium Enterprises) with cross-border trade intelligence, alternative credit scoring, supply chain optimization, and real-time market analytics.

Built for the BorneoHack hackathon, this full-stack application combines a FastAPI backend with a Next.js frontend, powered by Google Gemini AI and a LangGraph multi-agent orchestrator.

✨ Key Features

🤖 AI-Finance Assist (Multi-Agent Chatbot)

  • LangGraph Multi-Agent Orchestrator — Supervisor Agent delegates tasks to specialist sub-agents
  • Researcher Agent (7 tools) — Web search, ASEAN news, chart/graph generation, image generation, tariff guidance, TTS
  • Quant Agent (4 tools) — Calculator, stock prices, live currency conversion, loan amortization
  • Prompt Injection Protection — Hardened system prompts with security directives
  • MCP Integration — Model Context Protocol server exposing tools over SSE
  • Markdown-rendered responses with tables, charts, and images inline

🌐 Cross-Border Trade Navigator

  • AI-powered Q&A on ASEAN import/export regulations
  • RAG-based semantic search over trade regulation knowledge base (Supabase pgvector)
  • Auto-generated compliance documents (Commercial Invoice, Certificate of Origin)
  • Tariff lookup with HS code estimation under ATIGA framework

📊 Alternative Credit Scoring

  • AI credit model using 11+ non-traditional data points
  • Assesses unbanked/offline businesses via digital transaction ratios, mobile payment volume, inventory turnover, supplier reliability, and customer ratings
  • Visual score dial with risk tier classification and AI recommendations

📦 Automated Supply Chain Management

  • Smart supplier recommendation engine
  • Demand forecasting and inventory optimization
  • Regional supplier matching across ASEAN countries

🗺️ World Monitor Command Center

  • Real-time geopolitical risk surveillance with interactive map
  • Live USGS earthquake data layer integration
  • Multi-stream live news (Al Jazeera, Sky News, Euronews, DW)
  • Live webcam feeds from global hotspots
  • Country instability indices with dynamic risk scoring

🏪 Business Visibility Engine

  • Digital onboarding for offline businesses
  • Multi-step verified profile creation
  • Supply chain identity generation (MSME-ID)

📈 MSME Dashboard

  • Revenue & expense analytics with interactive Recharts visualizations
  • Smart Process Manager (Kanban-style workflow)
  • Market intelligence summary with regional breakdown

🏗️ Architecture

BorneoHack/
├── backend/                          # FastAPI Python Backend
│   ├── app/
│   │   ├── main.py                   # Application entry point & CORS
│   │   ├── config.py                 # Pydantic settings from .env
│   │   ├── routers/
│   │   │   ├── chatbot.py            # AI-Finance Assist endpoint
│   │   │   ├── trade_ai.py           # Trade Navigator (RAG + Compliance)
│   │   │   ├── credit.py             # Alternative Credit Scoring
│   │   │   ├── supply_chain.py       # Supply Chain Optimization
│   │   │   ├── market.py             # Predictive Market Analytics
│   │   │   ├── document_ai.py        # Document Processing
│   │   │   ├── translation.py        # Multi-language Translation
│   │   │   ├── inventory.py          # Inventory Management
│   │   │   ├── dashboard.py          # Dashboard Analytics
│   │   │   ├── mcp.py                # Model Context Protocol Server
│   │   │   └── health.py             # Health Check
│   │   ├── services/
│   │   │   ├── agent.py              # LangGraph Multi-Agent System (11 tools)
│   │   │   ├── trade_ai.py           # Trade AI RAG + Semantic Search
│   │   │   ├── gemini_client.py      # Google Gemini API Client
│   │   │   ├── credit.py             # Credit Scoring Engine
│   │   │   ├── supply_chain.py       # Supply Chain Logic
│   │   │   ├── market.py             # Market Analytics
│   │   │   ├── document_ai.py        # Document AI Service
│   │   │   ├── translation.py        # Translation Service
│   │   │   ├── inventory.py          # Inventory Service
│   │   │   ├── dashboard.py          # Dashboard Service
│   │   │   ├── mem0_client.py        # Mem0 Memory Client
│   │   │   └── supabase_client.py    # Supabase Client
│   │   └── schemas/                  # Pydantic request/response models
│   ├── requirements.txt
│   └── .env.example
│
├── frontend/                         # Next.js TypeScript Frontend
│   ├── src/
│   │   ├── app/
│   │   │   ├── page.tsx              # Main Dashboard
│   │   │   ├── trade-navigator/      # Trade Navigator Chat
│   │   │   ├── credit-scoring/       # Credit Scoring UI
│   │   │   ├── supply-chain/         # Supply Chain Management
│   │   │   ├── worldmonitor/         # World Monitor Command Center
│   │   │   ├── visibility-engine/    # Business Visibility Engine
│   │   │   └── globals.css           # Design System & Tokens
│   │   └── components/
│   │       ├── ai-chat.tsx           # AI-Finance Assist Chat Widget
│   │       ├── right-chat-sidebar.tsx# Sliding Chat Drawer
│   │       ├── sidebar.tsx           # Navigation Sidebar
│   │       └── world-map.tsx         # Interactive World Map
│   └── package.json
│
└── README.md

🧠 Multi-Agent System

The AI backbone uses LangGraph to orchestrate a Supervisor → Worker multi-agent pattern:

                    ┌─────────────┐
    User Query ───▶ │  Supervisor  │ ◀─── Structured Routing
                    │ (Orchestrator)│
                    └──────┬──────┘
                           │
              ┌────────────┼────────────┐
              ▼                         ▼
     ┌────────────────┐       ┌────────────────┐
     │   Researcher    │       │     Quant       │
     │   (7 Tools)     │       │   (4 Tools)     │
     ├────────────────┤       ├────────────────┤
     │ • search_web    │       │ • calculator    │
     │ • asean_news    │       │ • stock_price   │
     │ • current_date  │       │ • convert_curr  │
     │ • gen_image     │       │ • calc_loan     │
     │ • gen_graph     │       └────────────────┘
     │ • tariff_info   │
     │ • tts_url       │
     └────────────────┘

Security Features:

  • Prompt injection defense across all agent prompts
  • Domain-confinement (financial/trade topics only)
  • System instruction concealment directives
  • Input length capping on tool parameters

🚀 Getting Started

Prerequisites

  • Python 3.11+
  • Node.js 18+
  • npm or yarn

Backend (FastAPI)

cd backend

# Create & activate virtual environment
python -m venv venv
.\venv\Scripts\activate        # Windows
# source venv/bin/activate     # macOS / Linux

# Install dependencies
pip install -r requirements.txt

# Copy env and configure API keys
cp .env.example .env
# Edit .env with your API keys (see Environment Variables below)

# Run dev server
uvicorn app.main:app --reload --port 8000

API docs available at: http://localhost:8000/docs

Frontend (Next.js)

cd frontend

# Install dependencies
npm install --legacy-peer-deps

# Run dev server
npm run dev

App available at: http://localhost:3000

🔑 Environment Variables

Create a .env file in /backend with:

Variable Description Required
GEMINI_API_KEY Google Gemini API key ✅
GEMINI_MODEL Gemini model name (e.g. gemini-2.5-flash) ✅
TAVILY_API_KEY Tavily web search API key ✅
ALPHAVANTAGE_API_KEY Alpha Vantage stock/forex API key ⬚ Optional
NEWS_API_KEY NewsAPI.org key for ASEAN headlines ⬚ Optional
MEM0_API_KEY Mem0 memory layer API key ⬚ Optional
SUPABASE_URL Supabase project URL ⬚ Optional
SUPABASE_ANON_KEY Supabase anonymous key ⬚ Optional
SUPABASE_SERVICE_KEY Supabase service role key ⬚ Optional

📡 API Endpoints

Method Endpoint Description
GET / Root welcome message
GET /api/health Health check
POST /api/chatbot/message AI-Finance Assist (multi-agent)
POST /api/trade-ai/query Trade regulation Q&A (RAG)
POST /api/trade-ai/generate-document Compliance document generation
POST /api/trade-ai/tariff-lookup Tariff rate lookup
POST /api/credit/score Alternative credit scoring
POST /api/supply-chain/recommend-suppliers Supplier recommendations
GET /api/market/analysis Market analytics
POST /api/document-ai/process Document AI processing
POST /api/translation/translate Multi-language translation
GET /api/dashboard/summary Dashboard data
GET /api/mcp/ MCP Server (SSE)

🛠️ Tech Stack

Layer Technology
Frontend Next.js 16.1, React 19, TypeScript, Tailwind CSS v4
UI Components Lucide React, Recharts, react-simple-maps, react-markdown, remark-gfm
Backend FastAPI, Python 3.11+, Pydantic v2
AI/LLM Google Gemini (gemini-2.5-flash), LangChain, LangGraph
Agent Tools Tavily (search), Alpha Vantage (stocks), NewsAPI (news), QuickChart (graphs), Pollinations (images), ExchangeRate API (forex)
Database Supabase (PostgreSQL + pgvector)
Memory Mem0 (conversational memory)
Protocol Model Context Protocol (MCP) via FastMCP

🎥 Demo Video Link

https://drive.google.com/file/d/1lJO3Bpwct4cWzprqfkGMhjadmW7-dFuZ/view?usp=sharing

📄 License

This project is for hackathon demonstration purposes.

About

No description, website, or topics provided.

Resources

Readme
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages