AxelDlv00/EULAI-q4f16_1-MLC

You lie? EULAI!

Local AI Browser Assistant for Legal Document Analysis

Axel Delaval • 28 January 2026

EULAI - MLC/WebLLM Optimized Qwen3-0.6B for EULA Analysis

This repository contains the MLC-compiled version of AxelDlv00/EULAI. These files are optimized for browser-based inference using WebLLM (WebGPU) or local deployment via the MLC LLM runtime.

Model Description

EULAI is a fine-tuned version of Qwen/Qwen3-0.6B specialized in analyzing End-User License Agreements (EULA) and privacy policies. It extracts structured, risk-based summaries to help users understand legal implications quickly.

Files Available

• q4f16_1 quantization: 4-bit group quantization (4.5 bits per parameter). Optimized for a perfect balance between speed, weight (approx. 340 MB), and accuracy on WebGPU.
• MLC Config: Includes mlc-chat-config.json pre-configured for the ChatML conversation template.

Usage (Inference with WebLLM SDK)

To run this model in a web browser or a Chrome extension, you can use the @mlc-ai/web-llm package.

import * as webllm from "@mlc-ai/web-llm";

const modelConfig = {
  model: [https://huggingface.co/AxelDlv00/EULAI-q4f16_1-MLC", 
  model_id: "EULAI-q4f16_1-MLC",
  model_lib: "https://raw.githubusercontent.com/mlc-ai/binary-mlc-llm-libs/main/qwen3/Qwen3-0.6B-q4f16_1-webgpu.wasm" 
};

async function main() {
  const engine = await webllm.CreateMLCEngine(modelConfig.model_id, {
    appConfig: { model_list: [modelConfig] }
  });

  const prompt = "The application shares your browsing history with unspecified third parties.";
  const messages = [{ role: "user", content: prompt }];

  const reply = await engine.chat.completions.create({ messages });
  console.log(reply.choices[0].message.content);
}

Output Format

The model generates structured analysis using this strict format:

- [BLOCKER/BAD/GOOD/NEUTRAL] : Title : Explanation

Example

• Input: "The app shares your contact list with third-party advertisers."
• Output: - [BAD] : Data Sharing : Your personal contact information is shared with external parties for marketing purposes without explicit granular consent.

Technical Details

• Base Model: Qwen/Qwen3-0.6B
• Fine-tuning: PEFT / LoRA (Merged before MLC conversion)
• Quantization Format: q4f16_1 (4-bit, group_size=32)
• Target Runtime: TVM Unity / WebGPU
• Conversation Template: chatml

Disclaimer

This model is an AI assistant for legal document analysis and does not constitute professional legal advice. Always consult with a qualified lawyer for critical legal matters.