Personal finance agent: bank aggregation, spending analysis, and projections

[kovtcharov]

Summary

Personal finance agent built on GAIA's platform capabilities (RAG, DatabaseMixin, memory, Agent UI). The platform should enable this use case naturally — bank statement processing via RAG, transaction storage via DatabaseMixin, spending pattern learning via MemoryMixin, visualizations via observability dashboard.

This is both a reference agent demonstrating platform capabilities AND a genuinely useful consumer/small-business tool.

Why Local-Only Matters Here

Financial data is among the most sensitive personal information. Users will not (and should not) send bank statements, transaction histories, and spending patterns to cloud APIs. Local processing is the only acceptable architecture.

Key Features

1. Bank statement import — PDF/CSV import via existing RAG pipeline
2. Transaction categorization — AI-categorized spending via LLM
3. Spending analysis — trends, anomalies, month-over-month (via DatabaseMixin queries)
4. Budget tracking — set budgets per category, track against actuals
5. Financial projections — "what if" scenarios based on current patterns
6. Subscription detection — identify recurring charges, flag unused subscriptions
7. Multi-currency support — handle international accounts
8. Small business extension — invoicing, expense tracking, revenue analysis

Platform Capabilities Used

• RAG — parse bank statement PDFs
• DatabaseMixin — store and query transactions
• MemoryMixin — learn spending patterns over time
• Agent UI — observability dashboard for financial visualizations
• Autonomy engine — scheduled bill reminders, budget alerts

Acceptance Criteria

• Import bank statements (PDF and CSV) via existing RAG
• Auto-categorize transactions via LLM
• Monthly spending dashboard in Agent UI
• Budget tracking with alerts
• All financial data stored locally — zero cloud transmission
• Serves as reference implementation showing platform capabilities