Aza Man: CLI-Powered Autonomous Financial Agent

Aza Man is an AI-driven financial assistant built for the Agentic AI Innovation Challenge 2025. Operating via CLI, it autonomously manages budgets, tracks expenses, and optimizes savings using a tool-driven workflow and persistent state management.

Abstract

Financial management is daunting for many due to literacy gaps. Aza Man tackles this with a command-line interface that automates budgeting and expense tracking. Its agentic design leverages LangGraph for seamless, precise financial oversight.

Financial Management Challenges

Users often lack skills for budgeting and tracking expenses, relying on inconsistent guidance. Aza Man offers an autonomous, CLI-based solution, accessible to anyone with a terminal, eliminating UI complexity.

The Solution: Aza Man

Aza Man persists user data in SQLite, tied to a user ID, ensuring continuity across CLI sessions. It mandates budget setup before expenses, using tools like budget and log_expenses for accuracy. Its text-based interaction simplifies financial tasks.

Features

• Budget Setup: Enforces income, savings goal, and currency setup via budget.
• Expense Tracking: Logs expenses with log_expenses, updating totals.
• Insights: Calculates metrics (e.g., remaining budget) with math_tool.
• State Persistence: Retains data (username, income, expenses) across sessions.
• CLI Interface: Text-based, lightweight, and efficient.

Target Audience

Tech-savvy users, students, and professionals needing a no-frills, automated financial tool benefit from Aza Man’s CLI design.

Methodology

Built with LangChain and Python, Aza Man integrates a Google Gemini model via OpenRouter. Its LangGraph workflow (call_model, store_memory) ensures tool-driven autonomy, avoiding manual calculations.

System Design

• Architecture: Modular, with state management and tool logic.
• Tech Stack: Python, SQLite for storage, LangGraph for workflow.
• Workflow: Routes CLI inputs through call_model to tools, persisting updates.

Implementation

• Interface: CLI-based, handling inputs like "My username is John" or "exit".
• Data Management: SQLite stores user states.
• Logic: Processes financial tasks via tools, outputs results in terminal.

Testing & Debugging

• Unit Testing: Validated tool responses (e.g., set_username).
• Integration Testing: Ensured workflow coherence in CLI.
• User Testing: Refined based on terminal feedback.

Deployment & Maintenance

• Hosting: Runs locally via python main.py.
• Version Control: GitHub repository (link TBD).

Results

Aza Man works as intended in CLI, autonomously managing finances. Full deployment awaits final polish, but core functionality is robust.

Future Prospects

Future iterations may add UI support, currency localization, and data-driven insights, scaling its agentic potential.

Conclusion

Aza Man proves agentic AI’s value in financial automation. Its CLI simplicity and tool-driven precision make it a standout for efficient money management.

Tags: agentic-ai, finance, cli
Hub: #chinonsoodiaka
License: TBD

Table of Contents

• Abstract
• Financial Management Challenges
• The Solution: Aza Man
• Features
• Target Audience
• Methodology
• System Design
• Implementation
• Testing & Debugging
• Deployment & Maintenance
• Results
• Future Prospects
• Conclusion