Enterprise Autonomous Financial Intelligence Platform

An enterprise-grade, multi-agent AI system for autonomous financial document analysis, compliance checking, risk scoring, and portfolio intelligence — built for regulated financial environments.

Architecture Overview

Frontend (React + TypeScript + Vite)
        ↓  WebSocket + REST
FastAPI Gateway (Auth, Rate Limiting, Validation)
        ↓
LangGraph Orchestrator (Multi-Agent Brain)
        ↓
┌─────────────────────────────────────┐
│  Supervisor → Research → Risk       │
│  Compliance → Market → Strategy     │
│  Execution → Hallucination Guard    │
└─────────────────────────────────────┘
        ↓
Memory Layer: Redis + PostgreSQL + Qdrant
        ↓
LLMs + Financial APIs + SEC/SEBI Data
        ↓
Observability: LangSmith + Prometheus + Grafana

Key Features

• 7-Agent LangGraph Orchestration — Supervisor, Research, Risk, Compliance, Market, Strategy, Execution agents
• LLM Security Layer — Prompt injection detection, hallucination mitigation, confidence scoring
• Explainable Risk Scoring — SHAP-based feature attribution for every model decision
• Human-in-the-Loop Dashboard — Confidence-based escalation with full audit trail
• RAG over SEC/SEBI Filings — Hybrid search, BGE embeddings, reranking
• Real-time Anomaly Detection — Transaction stream monitoring with sliding window stats
• Multi-layer Memory — Redis (session), PostgreSQL (persistent), Qdrant (semantic)
• LangSmith + Prometheus Observability — Token usage, latency, hallucination tracking

Tech Stack

Backend

• FastAPI — Async API gateway
• LangGraph — Multi-agent orchestration
• LangChain — RAG, tools, memory
• PostgreSQL — Persistent storage
• Redis — Session/cache layer
• Qdrant — Vector store
• SHAP — Explainability
• Kafka / Redis Streams — Real-time event streaming
• LangSmith — Agent observability
• Prometheus + Grafana — Metrics

Frontend

• React 18 + TypeScript + Vite
• TailwindCSS + shadcn/ui
• Zustand — State management
• Recharts — Financial charts
• WebSockets — Real-time agent activity

Quick Start

# Clone
git clone https://github.com/VivekGaddam/enterprise-fin-ai
cd enterprise-fin-ai

# Backend
cd backend
cp .env.example .env   # Fill in API keys
pip install -r requirements.txt
alembic upgrade head
uvicorn app.main:app --reload

# Frontend
cd frontend
npm install
npm run dev

Environment Variables

# LLM
OPENAI_API_KEY=
ANTHROPIC_API_KEY=

# Database
DATABASE_URL=postgresql://user:pass@localhost:5432/finai
REDIS_URL=redis://localhost:6379
QDRANT_URL=http://localhost:6333

# Financial APIs
ALPHA_VANTAGE_KEY=
POLYGON_API_KEY=
FINNHUB_API_KEY=

# Observability
LANGCHAIN_API_KEY=
LANGCHAIN_TRACING_V2=true
LANGCHAIN_PROJECT=enterprise-fin-ai

# Security
SECRET_KEY=
JWT_ALGORITHM=HS256
ACCESS_TOKEN_EXPIRE_MINUTES=30

Project Structure

enterprise-fin-ai/
├── backend/
│   ├── app/
│   │   ├── api/v1/routes/       # FastAPI route handlers
│   │   ├── core/
│   │   │   ├── agents/          # LangGraph agent definitions
│   │   │   ├── memory/          # Redis, Qdrant, PG memory
│   │   │   ├── rag/             # RAG pipeline, embeddings
│   │   │   └── security/        # Prompt injection, auth
│   │   ├── models/              # SQLAlchemy ORM models
│   │   ├── schemas/             # Pydantic schemas
│   │   └── services/            # Business logic
│   ├── alembic/                 # DB migrations
│   └── requirements.txt
├── frontend/
│   ├── src/
│   │   ├── components/          # UI components
│   │   ├── pages/               # Route pages
│   │   ├── hooks/               # Custom React hooks
│   │   ├── store/               # Zustand stores
│   │   └── types/               # TypeScript types
│   └── package.json
└── docker-compose.yml