• Designed and built a question-answering system for unstructured spreadsheet data
• Developed a conversational agentic system using OpenAI-functions (GPT-3.5-Turbo) that reasons and acts based on either
lookup-type questions or analytical questions
• Implemented RAG-based RetrievalQA chain for direct answers from unstructured XLSX sheets for lookup questions
• Preprocessed & extracted tables for analytical questions to avoid LLM hallucination problems, then utilized a computing agent
that uses PythonAstREPLTool for accurate dataframe computations
Execute the following to install all requirements:
pip install -r requirements.txt
Execute the following command to start Ques-Ans system on the sheet
python sheetQA.py /path/to/file.xlsx openai_key