AI answers why i am broke

AI Answers: Why Am I Broke? 💸🤖

Ever wonder where your money goes every month? In 2026, we don't just wonder—we ask AI.

I built MoneyRAG, an AI-powered app that lets you CHAT with your bank statements. Upload your credit card CSVs and ask questions like:

"How much did I spend on fast food last year?"
"Am I paying for duplicate subscriptions?"
"Show me suspicious transactions"
🔗 Try it here: https://huggingface.co/spaces/SajilAw...
📂 GitHub: https://github.com/AwaleSajil/money-rag

⚙️ HOW IT WORKS (Technical Overview):

🔹 Streamlit UI - Upload CSV transaction files
🔹 LLM Schema Normalization - Handles any bank format (Chase, Discover, etc.)
🔹 Web Enrichment - DuckDuckGo search enriches confusing transaction descriptions
🔹 Dual Storage:
• SQLite for structured queries (math/aggregations)
• Qdrant vector DB for semantic search
🔹 MCP Server - Two tools:
• SQL Query Executor (precise calculations)
• Semantic Search (understanding intent)
🔹 LangChain Agent - Decides which tool to use based on your question
🔹 Stateful Memory - Remembers conversation context

🚀 GETTING STARTED:

Authenticate - Use Google Gemini or OpenAI API
• Get free Google API: https://aistudio.google.com/app/apikey
• Get OpenAI API: https://platform.openai.com/api-keys

Get Transaction History
• Chase Credit Card:    • ✅  How To Download Chase Transaction Histo...  
• Discover Credit Card:    • View Your Discover Credit Card Payment His...  

Upload & Ingest - AI does ETL processing, web enrichment, and vectorization

Ask Questions!
• "Where did I spend my money last month?"
• "How much on haircuts this year?"
• "Find all Starbucks purchases"

🎯 CHALLENGES SOLVED:

Long-tail Top-K Problem - How to set the right K value for semantic search?
Solution: Agent iteratively increases K if it detects incomplete results (long tail detected).

🔮 FUTURE DIRECTIONS:

📸 Bill Image Processing - Upload receipt photos, link to transactions
🍺 "How much did I spend on beer last year?"
🥗 "What food should I replace to eat healthier on a budget?"

🛠️ TECH STACK:

• LangChain & LangGraph (Agent orchestration)
• Google Gemini / OpenAI GPT-4 (LLMs)
• Qdrant (Vector database)
• SQLite (Structured storage)
• FastMCP (Model Context Protocol)
• Streamlit (Web interface)
• DuckDuckGo Search (Merchant enrichment)

👥 CONTRIBUTORS:

• Sajil Awale - https://github.com/AwaleSajil
• Simran KC - https://github.com/iamsims

#AI #MachineLearning #PersonalFinance #RAG #LLM #OpenSource #LangChain #Python #FinTech #DataScience