System Architecture

A modular, multi-layer architecture designed for scalable legal AI reasoning and real-time agent orchestration.

Architecture Overview

React Flow

Core Components

Case Parser Agent

Extracts structured issues, parties, jurisdiction, and relevant legal domains from raw case text.

Issue Extraction Module

Identifies and categorizes the core legal issues, applicable laws, and dispute dimensions.

Legal Research Module

Retrieves relevant precedents, statutes, and legal commentary via RAG from the knowledge base.

Plaintiff Agent

Constructs prosecution arguments with cited evidence and legal basis to support the plaintiff's position.

Defendant Agent

Builds defense arguments, challenges evidence, and presents counter-narratives for the defendant.

Rebuttal Agent

Generates rebuttals to each side's strongest arguments across multiple debate rounds.

Negotiation Agent

Explores settlement pathways and quantifies compromise positions between the parties.

Judge Agent

Evaluates all arguments, issues a structured judgment, and provides win probability estimates.

Technology Stack

Frontend

  • ▸Next.js 14 (App Router)
  • ▸TypeScript
  • ▸Tailwind CSS + shadcn/ui
  • ▸React Flow
  • ▸Framer Motion

Backend

  • ▸FastAPI (Python 3.11)
  • ▸LangGraph Orchestration
  • ▸Pydantic Schemas
  • ▸Uvicorn ASGI

Model Layer

  • ▸OpenAI GPT-4o / GPT-4
  • ▸Anthropic Claude
  • ▸Local LLM Support
  • ▸Function Calling

Knowledge Base

  • ▸Vector Store (Chroma/FAISS)
  • ▸Legal Document RAG
  • ▸Case Law Index
  • ▸Statute Database

Deployment

  • ▸Docker Compose
  • ▸Nginx Reverse Proxy
  • ▸Environment Configs
  • ▸Health Checks