FinanceOS Bot - AI Telegram bot for personal finance
Bot Development
Telegram bot for tracking expenses and income with AI analytics, natural language recognition, and automatic parsing of bank statements (Python, aiogram 3, Groq LLaMA, Supabase).
Main task: To free the user from manual financial management - the bot accepts messages in free form, automatically classifies transactions, tracks the budget, and provides AI recommendations based on spending history.
What has been implemented:
- Recognition of transactions from natural language (intent detection via LLaMA 3.3 70B) with automatic determination of amount, category, and type of operation
- Natural language parser for amounts - support for formats "25k", "half a million", "twenty thousand UAH"
- Parsing of bank CSV and PDF statements with automatic entry of transactions into the database
- Financial goals with deadlines, progress tracking, and savings plans
- AI financial advisor - answers to questions about expenses based on context from the database (embeddings + semantic search)
- Weekly automatic expense digest (APScheduler, cron)
- Onboarding of new users with FSM states
- Custom Supabase FSM storage for saving dialogue state between restarts
Technologies: Python, aiogram 3, Groq API (LLaMA 3.3 70B / 3.1 8B), LangChain, Supabase (PostgreSQL), sentence-transformers, APScheduler, pydantic-settings, loguru, Railway
Main task: To free the user from manual financial management - the bot accepts messages in free form, automatically classifies transactions, tracks the budget, and provides AI recommendations based on spending history.
What has been implemented:
- Recognition of transactions from natural language (intent detection via LLaMA 3.3 70B) with automatic determination of amount, category, and type of operation
- Natural language parser for amounts - support for formats "25k", "half a million", "twenty thousand UAH"
- Parsing of bank CSV and PDF statements with automatic entry of transactions into the database
- Financial goals with deadlines, progress tracking, and savings plans
- AI financial advisor - answers to questions about expenses based on context from the database (embeddings + semantic search)
- Weekly automatic expense digest (APScheduler, cron)
- Onboarding of new users with FSM states
- Custom Supabase FSM storage for saving dialogue state between restarts
Technologies: Python, aiogram 3, Groq API (LLaMA 3.3 70B / 3.1 8B), LangChain, Supabase (PostgreSQL), sentence-transformers, APScheduler, pydantic-settings, loguru, Railway