Crypto Accounting Chatbot

A standalone Streamlit chatbot app for crypto accounting, featuring CSV upload, DeepSeek R1 (Ollama) LLM integration, ChromaDB vector store, and a friendly, professional Penny persona. Penny can analyze uploaded CSVs, answer accounting questions, and route data to csv2api as needed... all with structured logging and persistent chat history.

Features

• Conversational Chat UI with persistent history and Penny persona
• CSV Upload & Preview for crypto accounting data
• Local LLM Integration (DeepSeek R1 via Ollama)
• ChromaDB Vector Store for persistent chat and CSV content storage
• ChromaDB Visualizer for exploring stored vectors
• Structured Logging of all user actions, routing, and LLM responses
• csv2api Integration: Calls csv2api as a subprocess for advanced CSV analysis (no code dependency)
• Intent Detection and prompt engineering for accurate, persona-driven responses
• Test Scripts for ChromaDB and logger sanity checks

How to Run

1. Install dependencies:

   pip install -r requirements.txt

2. Ensure Ollama and DeepSeek R1 are running locally.
3. Start the app:

   streamlit run app.py

4. (Optional) Visualize ChromaDB contents:

   streamlit run visualize_chromadb.py

5. (Optional) Run tests:

   python test_chromadb.py
   python test_logger.py

Architecture

• app.py: Main Streamlit app (chat UI, CSV upload, LLM/csv2api integration, logging)
• backend/csv_handler.py: Persona prompt engineering, LLM chat, intent detection, sync csv2api runner
• backend/vector_store.py: ChromaDB vector store logic
• utils/logger.py: Custom structured logger
• visualize_chromadb.py: ChromaDB visualizer UI
• csv2api/: Submodule, called as a subprocess (no direct code dependency)

Notes

• This app is fully independent of csv2api logic, but can call csv2api as a subprocess for advanced CSV analysis.
• All LLM chat and intent detection use the Penny persona prompt for consistent, professional responses.
• All major requirements (persona, sync/async, logging, ChromaDB, csv2api routing) are implemented.

---

For crypto accounting teams and professionals seeking a robust, privacy-friendly, and extensible AI assistant for CSV-based workflows.