Risk-Analyzer

VisaCompanion is a local tool that helps legal professionals and EB-1A petitioners identify weaknesses in their immigration petition drafts before submission. It simulates a USCIS adjudicator’s feedback using a lightweight open-source language model mistral running locally via Ollama.

---

Features

• AI-powered analysis of EB-1A petition content
• Section-by-section feedback modeled after USCIS RFE criteria
• 📄 Supports .docx, .pdf, and .txt files
• Runs entirely offline — no internet or API keys required

---

Folder Structure

Risk-Analyzer/
├── main.py # Analyzer pipeline
├── scraper.py # Downloads USCIS, AAO, Reddit data
├── sample_data/ # Petition input file(s)
├── outputs/ # Final Word report + chart
├── knowledge_base/
│ ├── raw/aao/ # Downloaded AAO decision PDFs
│ └── processed/ # Extracted JSON knowledge
├── src/
│ ├── parser.py
│ ├── risk_detector.py
│ └── report_generator.py
├── requirements.txt
└── README.md

---

Requirements

• Python 3.8 or higher
• Ollama (to run the LLM locally)
• At least 8 GB of RAM
• Microsoft Word or Google Docs (to open output files)

---

Setup Instructions

Build the Knowledge Base

Before you run the analyzer, you must scrape the official EB-1A resources into the local knowledge base.

python scraper.py

1. Install Ollama

Download the installer from:
https://ollama.com/download
Install and restart your terminal (or PowerShell) afterward.

---

2. Pull the Mistral LLM

In your terminal or PowerShell, run:

ollama pull mistral

To start the model, open a new terminal window and run:

ollama run mistral

Keep this terminal running. It hosts the LLM in the background while your tool runs.

---

3. Install Python Dependencies

In your main project folder, run:

pip install -r requirements.txt

This will install all dependencies .

---

How to Use the Tool

Step 1: Add Your Petition Draft

Save your .docx or .pdf petition draft inside the sample_data/ folder.

Step 2: Run the Script

In your terminal, from the root of the project, run:

python main.py

This will take a whileee

Step 3: Review the Results

After completion, the feedback report will be saved at:

outputs/rfe_risk_report.docx

Open it in Microsoft Word or Google Docs to view the analysis.

---

What the Output Includes

Each analyzed section contains:

• The EB-1A criterion name
• Your original petition content
• AI-generated insights:
  • Weaknesses in evidence or phrasing
  • Missing elements that may trigger RFE
  • Suggestions to strengthen the section

Credits

• Ollama
• Mistral
• USCIS EB-1A Policy Manual