bennedictbett/Mpesa-Adviser-Financial-

cat > /mnt/user-data/outputs/README.md A RAG-powered personal finance coach for Kenyans. Upload your M-Pesa statement PDF and ask natural language questions about your spending, saving, and sending habits. Answers are cited directly from your real transaction data — no guessing, no generic advice. **Live demo:** mpesa-adviser-financial.vercel.app --- ## The Problem Kenya has 30M+ active M-Pesa users. Yet most Kenyans have no clear picture of their financial health. Your M-Pesa statement is a PDF sitting in your email — full of valuable data about your spending, saving, and sending habits — but nobody reads it. Most people have no idea how much they spend on food, transport, or airtime every month, making it impossible to budget, save, or make better financial decisions. General-purpose AI chatbots make this worse by answering financial questions from outdated training data with no source to verify against. --- ## The Solution Upload your M-Pesa statement and ask: - *"How much did I spend last month?"* - *"What do I spend most of my money on?"* - *"How can I save KES 5,000 this month?"* - *"How much did I send to family in April?"* - *"How much does it cost to send KES 2,500?"* The advisor reads your actual transaction history, categorises every transaction, and gives you grounded, practical advice cited directly from your own data and official Safaricom/CBK documents. --- ## Architecture ``` ┌─────────────────────────────────────────────────────────┐ │ INGESTION PIPELINE │ │ (runs once at setup) │ │ │ │ PDF docs → pdf_parser → splitter → embeddings │ │ ↓ │ │ ChromaDB │ │ (vector store) │ └─────────────────────────────────────────────────────────┘ ┌────────────────────── …