Not-A-financial-advisor-for-legal-reasonsBot

Inspiration The idea for this project came from the frustration of managing personal finances and keeping track of spending. Many people struggle with understanding their financial habits or need quick insights into their expenses. As a developer, I wanted to create a financial assistant that could help users quickly get answers to their financial questions, all while keeping things user-friendly and intuitive. However, due to the complexities and regulations around financial advice, I designed this chatbot as an informational tool only, hence the name, Not-A-financial-advisor-for-legal-reasonsBot.

What it does Not-A-financial-advisor-for-legal-reasonsBot is an AI-powered financial assistant that helps users retrieve insights about their spending and expenses. Users can ask questions like:

"How much did I spend on rent in the last month?" "What were my biggest purchases last week?" "How much have I spent on coffee this year?" The bot is capable of pulling this data from a user's payment history and providing quick, actionable insights without any technical complexity.

How we built it We built the bot using the following tools and technologies:

Streamlit for the user interface, making it easy to deploy the assistant as a web application. RAG (Retrieval Augmented Generation) chain for invoking the AI model and querying the payment data, ensuring that responses are not only conversational but also accurate. Paypal API and canned document vector stores for storing and querying payment history, ensuring fast and efficient retrieval of financial records. Python for backend logic and connecting all components together. The RAG model leverages the GPT-based architecture to generate natural language responses, while

Challenges we ran into One of the main challenges was managing user data in a way that maintained performance and privacy. Storing financial records is sensitive, so we had to ensure that no actual financial advice is provided, just insights into past spending habits.

Another challenge was optimizing the chatbot's performance, as real-time querying and generating responses based on large datasets can become slow. We focused on reducing latency and making the system as responsive as possible, especially for frequent users with a large number of transactions.

Accomplishments that we're proud of Successfully integrating RAG with financial data to provide accurate and quick insights from user queries. Building a user-friendly interface using Streamlit that makes financial data easy to access and understand. Maintaining user privacy and legal boundaries by ensuring the bot doesn’t give direct financial advice. Optimizing performance to handle large datasets with minimal lag. What we learned This project taught us a lot about the complexities of working with financial data, especially around legal boundaries and privacy concerns. We also deepened our understanding of integrating AI systems like RAG with traditional databases (such as AlloyDB), as well as balancing performance with user experience.

We also learned that simplifying complex systems (like financial data retrieval) into an easy-to-use chatbot interface requires careful attention to both technical details and the overall user journey.

What's next for Not-A-financial-advisor-for-legal-reasonsBot Advanced Insights: We plan to introduce more advanced features such as automatic monthly spending reports, trend analysis, and personalized financial tips (still not financial advice!). Multi-language Support: Expanding the assistant’s capabilities to support multiple languages, helping users from diverse backgrounds. Mobile Integration: Developing a mobile app version of the bot so users can access their financial insights on the go. Collaboration with Financial Apps: Integrating the bot with popular financial apps like Mint or Plaid to provide even more personalized data and help users better manage their finances.

Built With

• langchain
• perplexity
• python
• streamlit