🧠 Consulta AI Immigration Assistant

A working prototype of an AI-powered assistant that answers U.S. immigration-related queries using Retrieval-Augmented Generation (RAG). Built with LangChain, FAISS, and Streamlit, the assistant helps users get accurate, grounded responses from official sources like USCIS.

⚡ Designed in alignment with Consulta's mission to merge law, automation, and AI for impactful immigration solutions.

• 🔍 Query-Answering: Ask natural language questions about immigration
• 📚 Knowledge Base: Uses USCIS public content (scraped or uploaded)
• 🧠 RAG Architecture: Retrieval + LLM generation pipeline
• 🛠️ LangChain + FAISS: Efficient vector search and retrieval
• 🌐 Streamlit Interface: Interactive and deployable UI

(Insert screenshots or a Loom demo link here if available)

1. Clone the Repository bash Copy Edit git clone https://github.com/rxl895/consulta-ai-immigration-assistant.git cd consulta-ai-immigration-assistant
2. Install Python Dependencies bash Copy Edit pip install -r requirements.txt
3. Set Environment Variables Create a .env file:

bash Copy Edit touch .env Paste your Hugging Face API token inside:

env Copy Edit HUGGINGFACEHUB_API_TOKEN=your_huggingface_token_here 🔐 Never commit this file. It’s already included in .gitignore.

4. Run the Full Pipeline Step 1: Scrape USCIS Content bash Copy Edit python3 src/ingest.py Step 2: Embed the Content with Sentence Transformers bash Copy Edit python3 src/embed.py Step 3: Run the Chatbot Interface bash Copy Edit streamlit run src/app.py Visit the generated localhost link to chat with your AI Assistant.

“How to apply for an H1B visa?”

“What documents do I need for OPT?”

“What is I-130 used for?”

“How to change visa status inside the U.S.?”

LangChain for chaining and context injection

FAISS for similarity search over USCIS content

HuggingFaceHub for hosted open LLMs (e.g. Zephyr)

Streamlit for interactive front-end

BeautifulSoup for scraping USCIS.gov content

Add citation links for retrieved USCIS content

Multilingual support (Spanish, Hindi)

Add user feedback for answer helpfulness

Public deployment on HuggingFace Spaces

Add token usage monitoring and cost display

Created with ❤️ by Ritika Lamba Inspired by Consulta’s mission to revolutionize immigration

consulta-ai-immigration-assistant/
├── data/                        # Stores scraped content & FAISS index
│   ├── uscis_content.txt        # Raw USCIS content
│   └── uscis_faiss_index/       # Saved FAISS vector database
├── docs/                        # Assets for README/demo
│   └── README_assets/           # Screenshots or GIFs
├── src/
│   ├── app.py                   # Streamlit frontend & QA pipeline
│   ├── ingest.py                # Web scraper for USCIS site
│   └── embed.py                 # Embeds scraped data into FAISS index
├── .env                         # Hugging Face / OpenAI API keys (not committed)
├── requirements.txt             # Python dependencies
└── README.md                    # Project overview