README.md

metadata

title: Precision Legal Intelligence AdeemAlotaibi
emoji: ⚖️
colorFrom: gray
colorTo: red
sdk: docker
pinned: false

⚖️ LEX - Legal Expert Agentic System

LEX is an AI-powered legal analysis system designed to bridge the "Access to Justice Gap". It helps users navigate intimidating legal documents and proceedings by providing immediate clarity through a sophisticated multi-agent workflow. +4

🏗️ System Architecture: Optimized Parallel Pipeline

LEX utilizes an AISA (Agentic Interactive System Architecture) workflow that runs 6 specialized agents across 4 logical phases. To ensure a professional and responsive user experience, the system employs Python's ThreadPoolExecutor to reduce processing time from ~20 seconds to just 10–12 seconds.

🚀 Setup & Installation

• Prerequisites Python 3.9+ OpenAI API Key (Optimized for gpt-4o-mini) Required Libraries: pydantic, faiss-cpu, sqlite3, concurrent.futures

• Installation Clone the repository: Bash: git clone https://huggingface.co/spaces/AISA-Framework/LEX-AdeemAlotaibi cd LEX-AdeemAlotaibi

Install dependencies: Bash: pip install -r requirements.txt

Configure Environment: Set AISA_DETERMINISTIC=1 for reproducible test results

🛠️ Usage

LEX is designed to be intuitive for users facing legal stress.The system processes messy, real-world text and converts it into a structured legal analysis.

1. Inputting a Case

Users can input their legal situation in the text area (up to 8,000 characters).This can include:

• Narratives: Describing a dispute or situation in plain language.
• Document Text: Pasting the content of a termination notice, contract clause, or court judgment.
• Multilingual Input: Input can be in Arabic or English; the system is robust enough to categorize and analyze both.

2. The Analysis Process

Once submitted, the system triggers the AISA Workflow:

• Instant Identification: The system immediately identifies the "Intent" (e.g., is this a notice, a contract, or a judgment?).
• Plain Language Explanation: It translates complex legal jargon into an easy-to-understand summary using the RAG tool.
• Automated Quality Check: The Critic Agent reviews the analysis. If it misses a key legal point (like a specific 90-day notice rule), it automatically re-runs the logic to ensure accuracy.

3. Understanding the Output

The user receives an Analysis Report containing:

• Summary: A clear explanation of "what is happening."
• Risk Assessment: A visual indicator of the risk level (Low, Medium, or High).
• Action Plan: Targeted next steps (e.g., "mentions notice period violation" or "first action step urgency=high").
• Key Points: Critical legal facts detected by the Reasoning Agent.

4. Example Scenarios

Scenario	Expected Output
Landlord Notice	Intent: Notice, Risk: Medium, Action: Notice period validation.
Salary Reduction	Intent: Contract, Risk: Medium, Action: Wage violation check.
Court Judgment	Intent: Judgment, Risk: High, Action: Urgent payment/appeal steps.

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference