DEV Community

Sk Farhad
Sk Farhad

Posted on

Building an Agentic Personal Financial Assistant

πŸš€ Building ChatExpense: An AI-Powered Expense Tracker That Actually Understands You
"What if tracking expenses was as easy as chatting with a friend?"
That’s the question I asked myself before starting ChatExpense β€” a smart personal finance assistant that understands natural language, receipts, and even voice inputs.

πŸ’‘ The Problem
Managing expenses is boring. Logging every coffee, bill, or grocery item takes time β€” and most apps out there feel like fancy spreadsheets with good UI. I wanted something conversational. Something... agentic.

πŸ€– Enter: ChatExpense

Image description

ChatExpense lets you:

  • Log expenses via chat, voice, or receipt photos
  • Auto-categorize and generate reports
  • Ask questions like β€œHow much did I spend on food last week?” or β€œWhat’s my biggest spending habit?”
  • Get personalized tips on saving, budgeting, and patterns

It’s not just rule-based β€” it's powered by OpenAI, Whisper, and a LangGraph-style agentic workflow running on top of Django REST + PostgreSQL.

πŸŽ₯ Demo Video (Watch it in action!)
πŸ‘‰ Watch the full feature demo: https://youtu.be/oOmMewFYTEA

This demo shows voice-based input, photo upload, chat queries, smart categories, and AI responses.

πŸ› οΈ Tech Stack:

  • Backend: Django, Django REST Framework, async APIs
  • Frontend: Streamlit (soon switching to Gradio/NextJS)
  • LLM Engine: OpenAI (GPT-4o)
  • Voice: OpenAI Whisper
  • Image Parsing: OCR pipeline for receipt understanding
  • Data Storage: PostgreSQL
  • Deployment: Railway, with a cost-efficient container split

πŸ” Challenges I Faced:

  • Managing cost of AI inference with OpenAI usage-based billing
  • Async Django APIs that play well with LLM back-and-forth messaging
  • Token security (especially for chat sessions and user-specific data)
  • Finding a way to merge chat UX with structured finance data

πŸ‘·πŸ½ Build in Public: I Need Your Feedback!
ChatExpense is still in testing, and I’m giving 2 months of Power User access to testers who use it and send feedback.

πŸ“² Test Android App:
https://play.google.com/apps/testing/com.chatexpense.app

✍️ Sign up to become a tester:
https://forms.gle/dMcGKrCW4QKF6Vz76

πŸ’¬ Join the Feedback Group:
https://www.facebook.com/groups/1035231038196557

🌐 Visit Website:
https://chatexpense.com

πŸ™ How You Can Help

  • If this sounds useful, I’d love if you:
  • Try the app for a few days
  • Send feedback (even one line)
  • Share any bugs, UI quirks, or features you’d love to see

Also, if you’ve worked with AI agents or expense tracking β€” I’d love to hear how you tackled data structuring and context management. Let’s nerd out in the comments πŸ‘‡

❀️ Let’s Make Expense Tracking Fun Again
Thanks for reading. I’m building ChatExpense to be helpful, not annoying.
Let’s turn boring budgeting into a conversation you actually enjoy.

buildinpublic #django #openai #saas #personalfinance #chatbots #langchain #aiagents

Top comments (0)