A Resilient Financial Routing Agent

• 

  Error

• 

  Code writing

• 

  Final

Inspiration

​SmartSpend AI is a resilient, autonomous transaction routing agent designed to eliminate manual data entry in personal finance. By replacing rigid tracking forms with human-centric natural language processing, it transforms casual conversational text into structured ledger data instantly.

​#What it Does ​The agent acts as an intelligent financial buffer. It continuously listens for casual user expense descriptions, extracts multi-entity costs, automatically maps items to correct budget pools using token logic, and handles over-budget state management seamlessly without breaking or crashing.

​#How We Developed It ​The system architecture was constructed entirely using Python. We utilized regular expressions (re) for dynamic numerical token extraction, built a semantic keyword-boundary matrix for category routing, and engineered state-variable tracking pools to handle mathematical balances locally.

​#Challenges We Ran Into ​Multi-Entity Splitting: Sentences with compound tracking requests (like spending on travel and snacks simultaneously) initially caused token drops. We resolved this by building a logical sharding loop using conjunction boundaries. ​Focus Drift: Minimizing UI friction required overriding terminal execution orders to lock typing focus onto the input prompt immediately upon boot.

​#Accomplishments That We Are Proud Of ​We successfully engineered a Cross-Category Overflow Management Matrix. When a specific category pool (like Transport) is completely exhausted, the agent doesn't throw a negative value error. Instead, it fluidly drains the targeted pool to zero and routes the remaining overflow debt directly from a secondary pool (like Food) in real-time.

​#What We Learned ​We gained deep insight into string tokenization, persistent state math calculations, and user-experience optimization. We also learned how to handle environment discrepancies by transitioning from a standard interface layout into a stable, high-availability terminal execution loop.

​#What's Next for SmartSpend AI ​We plan to integrate a true Machine Learning LLM API for advanced semantic understanding, hook the backend up to a live SQL database for permanent storage, and implement predictive analytics to alert users on spending patterns before their budgets overflow.

Built With

• cil
• git
• github
• os-level-terminal-focus-hooks
• python
• token-based-string-parsing-logic
• utilizing-native-data-structures-for-in-memory-ledger-state-management