LexiSnap

Real-time market intelligence from domain-specific news with sentiment analysis and ETF charts.

Features

• 4 domains: Business & Finance, Law & Government, Politics, Technology
• Market-moving filter: Keywords + Ollama Mistral to surface relevant news
• Domain sentiment model: Multi-domain RoBERTa adapter for finance/tech/law/politics
• ETF charts: SPY, QQQ, TLT, GLD (last 10 days)

Setup

1. Install dependencies

pip install -r requirements.txt

2. Environment variables

Copy .env.example to .env and add your API keys:

• SERPAPI_API_KEY – serpapi.com (for Google Trends)
• NEWS_API_KEY – newsapi.org

3. Train the domain sentiment model

The model file (domain_sentiment_model.pt) is too large for GitHub. Train it locally:

python domain_updated.py

This creates domain_sentiment_model.pt in the project root.

4. (Optional) Ollama for summarization

For Mistral summaries, run Ollama with Mistral:

ollama pull mistral
ollama serve

Run

./run_app.sh

Or:

streamlit run app.py

Project structure

• app.py – Streamlit app
• data_sources.py – News API, filters, Ollama summarization
• domain_updated.py – Domain sentiment model (train & inference)
• domain_sentiment_model.pt – Trained model (create via step 3)