[zeda0104]

Summary

This PR produces a complete working sample demonstrating hyper-personalization in banking using Dozer, Vector Databases, and LangChain, fulfilling the requirements described in #1690 and the Dozer architecture article.

What is included:

1. Multi-Source Banking Datasets (examples/llm-banking/data/):
   • customer_profiles.csv: Real-world demographics, credit scores, account tiers, and initial preferences.
   • transactions.csv: Transaction logs across categories (Travel, Dining, Groceries, Shopping, etc.).
   • credit_card_products.csv: Credit card product catalog with interest rates, annual fees, and detailed perks.
2. Dozer SQL Streaming Configuration (dozer-config.yaml):
   • Ingests multiple CSV tables into real-time aggregates.
   • Computes customer spending totals by category and exposes clean REST/gRPC endpoints (/customer_spending_summary).
3. Vector RAG & LangChain Recommendation Pipeline (recommend.py, vector_rag.py):
   • Embeds product perks into a semantic vector store.
   • Matches real-time customer spending profiles against optimal card benefits.
   • LangChain prompt synthesizer explains exactly why the recommended card fits the user's spending habits.
4. Interactive CLI & Docker Environment (app.py, docker-compose.yml):
   • python app.py CUST001 provides instant offline demo outputs.
   • docker-compose.yml provides containerized orchestration.
5. Comprehensive Unit Tests (tests/):
   • 11 unit tests covering data schema, Dozer YAML config, vector similarity search, and recommendation outputs.
   • 100% test pass rate with zero external API dependencies.

Fixes #1690

[CLAassistant]

All committers have signed the CLA.