Spaces:
Sleeping
Sleeping
| 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 | |