AI Data Visualization Agent

Non-technical users (business analysts, researchers, students) struggle to explore and visualize data because they must write code (SQL, Python, ggplot) or know visualization grammars. There is no easy natural-language-to-insight tool that lets users simply ask questions like 'Show me trends by month' and get accurate charts and explanations.

A local interactive AI assistant that takes natural language queries about datasets, generates appropriate visualizations (plots, charts, tables), and provides narrative insights and explanations. The system runs locally using open-source models and code execution, making data exploration accessible without coding.

• Upload a dataset
• Ask a question in natural language
• The system interprets the query and generates analysis steps
• Code is generated and executed in a sandbox
• Charts and narrative insights are returned

• Query interpreter — translates natural language into structured analysis steps
• Code generator — produces Python visualization code (Altair, Matplotlib)
• Execution orchestrator — runs code safely in a sandbox and returns results
• Narrator — creates human-friendly text insights, not just charts

• Frontend: Streamlit
• Backend: Python
• LLM: Ollama + local models (Mistral, Llama 3)
• Sandbox: Pydantic, subprocess
• Plots: Altair, Matplotlib, Seaborn
• Optional: RAG / memory for follow-up questions