The Bottleneck in Modern BI
Business intelligence is often bottlenecked by the technical gap between business users and data tools. When a stakeholder needs to know "What were our top-performing regions last quarter?", they usually have to wait for a data team to build a dashboard or pull a custom report.
While GenAI promises to solve this, standard LLMs introduce two massive dealbreakers for enterprise data: they hallucinate math, and they pose severe data privacy risks.
I built PROJECT_AGENTIC_BI to solve this exact problem.
Tools and Services
- Streamlit – Serves as the interactive, conversational frontend UI.
- ngrok – Provides a secure network tunnel, exposing the local n8n webhook to seamlessly intercept frontend requests.
- n8n (via Docker) – The backend orchestrator that manages the agentic workflow and API routing.
- Google Gemini 2.5 Flash – The LLM engine that translates natural language into strict JSON visual blueprints.
- Python & Pandas – The local deterministic engine that executes all data filtering and math to prevent AI hallucinations.
- Plotly – Renders the dynamic, native visualizations.