Legal Adaptive Routing Framework (LARF)

An Agentic AI Framework for Processing Philippine-Hong Kong Migrant Workers Legal Queries

Saint Louis University Baguio | Team 404FoundUs

Documentation • Report Bug • Request Feature

---

Overview

The Legal Adaptive Routing Framework (LARF) is a specialized Python framework designed to bridge the gap between informal user queries (often in Taglish) and formal legal reasoning. It employs a multi-stage Agentic Pipeline to intelligently process, route, and resolve legal queries with high accuracy.

The Agentic Pipeline

graph LR
    A[User Query<br/>Taglish/Tagalog] --> B(1. Normalize);
    B --> C(2. Classify);
    C --> D(3. Retrieve);
    D --> E(4. Generate);
    E --> F(5. Audit);
    F --> G[Legally Grounded<br/>Response];
    
    classDef step fill:#f9f9f9,stroke:#333,stroke-width:2px;
    class B,C,D,E,F step;
    style A fill:#e1f5fe,stroke:#0288d1,stroke-width:2px
    style G fill:#e8f5e9,stroke:#388e3c,stroke-width:2px

Loading

1. Normalize: Translate linguistic variations (Taglish/Tagalog) into standard legal English.
2. Classify: Intelligently route queries to the correct domain (e.g., General Information vs. Complex Reasoning).
3. Retrieve: Advanced RAG (Retrieval-Augmented Generation) search mechanism focused on semantic querying of specific jurisdictional indices (e.g., Philippine and Hong Kong Legal Statutes).
4. Generate: Produce legally grounded responses using specialized LLMs.
5. Audit: Validate generated output for safety and minimize hallucinations.

---

Features

• Multi-lingual Support: Native handling of Tagalog and Taglish inputs.
• Smart Routing: Directs queries to the most appropriate legal index or reasoning engine.
• Modular Architecture: Built for scalability and easy integration into existing systems.
• Robust Fact-Checking: Built-in mechanisms to reduce AI hallucinations in legal contexts.

---

Project Structure

Click to expand the directory structure

LegalAdaptiveRoutingFramework/
├── src/
│   └── adaptive_routing/
│       ├── config.py           # Global Configuration
│       ├── core/               # Low-level Engine
│       │   ├── engine.py       # OpenRouter API Handler
│       │   └── exceptions.py   # Custom Errors
│       └── modules/
│           ├── multihead_classifier/   # Triage Components
│           │   ├── detector.py
│           │   └── linguistic.py
│           ├── semantic_router/        # Routing Components
│           │   ├── legal_generation.py
│           │   └── logic_classifier.py
│           ├── legal_retrieval/        # RAG Components
│           │   ├── embedding.py
│           │   └── retriever.py
│           ├── safety_audit/           # Safety Components
│           │   ├── response_audit.py
│           │   └── safety_audit.md     # Module Documentation
│           ├── retrieval.py    # Legal Retrieval Facade
│           ├── router.py       # Router Facade
│           └── triage.py       # Triage Facade
├── tests/                      # Unit Tests
├── docs/                       # Documentation
├── main.py                     # CLI Driver Script
├── requirements.txt            # Python Dependencies
└── .env                        # Secrets (Excluded from Git)

---

⚡ Quick Start

Prerequisites

• Python: 3.10 or higher
• API Key: OpenRouter API Key for LLM access

Installation & Setup

1. Clone the repository

   git clone https://github.com/SLU-404FoundUs/Legal-Adaptive-Routing-Framework.git
   cd Legal-Adaptive-Routing-Framework

2. Setup Virtual Environment

   Choose the instructions for your operating system:

   Windows (PowerShell)

   # 1. Set execution policy (Required to run scripts)
   Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope CurrentUser
   
   # 2. Create virtual environment
   python -m venv myvenv
   
   # 3. Activate virtual environment
   .\myvenv\Scripts\Activate.ps1
   
   # 4. Install dependencies
   pip install -r requirements.txt

   macOS / Linux (Terminal)

   # 1. Create virtual environment
   python3 -m venv myvenv
   
   # 2. Activate virtual environment
   source myvenv/bin/activate
   
   # 3. Install dependencies
   pip install -r requirements.txt

3. Initialize and Run The .env configuration file is automatically generated when you first run either the CLI or Web interface.

   [!TIP] To utilize the latest stable configuration for both WEB.py and CLI.py, it is highly recommended to import the stable_veritas_v1.config file (located in localfiles/) via the Configuration menu in the Web Interface.

   Run CLI Version:

   python CLI.py

   Run Web Version:

   python WEB.py

   Note: On your first run, you will be prompted to enter your OpenRouter API key and preferred models.

---

Usage

CLI Version

Launch the interactive command-line interface with customizable settings and an isolated terminal popup:

python CLI.py

Note: A discrete popup terminal will launch to run the assistant clearly on both Windows and macOS.

GUI Version

To use the Web-based Graphical User Interface, start the Flask web server:

python WEB.py

Then, open your browser and navigate to the address displayed in the terminal (usually http://localhost:5220).

Using as a Library

You can import the modules directly into your Python application:

from src.adaptive_routing import TriageModule, SemanticRouterModule

# 1. Initialize Modules
triage = TriageModule()
router = SemanticRouterModule()

# 2. Process Input (Taglish -> English)
input_text = "Tinanggal ako sa trabaho ng walang notice."
result = triage._process_request_(input_text)
normalized_text = result['normalized_text'] 
# Output: "I was terminated from my job without notice."

# 3. Route & Generate Legal Response
if normalized_text:
    response = router._process_routing_(normalized_text)
    print(f"Advice: {response['response_text']}")

---

Documentation

For detailed API references, configuration options, and architectural diagrams, please refer to the Full Documentation.

---

⚠️ AI Disclaimer

Portions of the code in this framework were generated with the assistance of an agentic AI. However, all components have been thoroughly tested, reviewed, and validated by the developers to ensure safety, correctness, and adherence to legal logic standards.

---

Meet the Team (404FoundUs)

• Deleon, Earl Macy — 2221816@slu.edu.ph
• Diola, Josh Mckenzie — 2225962@slu.edu.ph
• Lachica, Rafael — 2195465@slu.edu.ph
• Lucban, Prince John Louie — 2225254@slu.edu.ph
• Navarro, Josiah Ezra — 2233059@slu.edu.ph
• Ramos, Albert Jannsen — 2221023@slu.edu.ph
• Retuta, Ian Benedick — 2223041@slu.edu.ph
• Yuen, Ka Hang Christian — 2214959@slu.edu.ph

---

Contribution

Contributions are welcome! Please ensure that you follow the Technical Documentation Standards when adding new modules.

1. Fork the Project
2. Create your Feature Branch (git checkout -b feature/AmazingFeature)
3. Commit your Changes (git commit -m 'Add some AmazingFeature')
4. Push to the Branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

---

📜 License

Distributed under the MIT License. See LICENSE for more information.

_{Thesis Study by Team 404FoundUs}