Intent-to-Analytics Agentic System

Modern business teams rely heavily on data but face a critical bottleneck like SQL dependency or even basic python analysis. - Non-technical users (product, marketing, ops) cannot query databases directly or run python analysis - Data teams are overloaded with repetitive query requests - Existing BI tools require rigid dashboards and lack flexibility - LLM-based solutions often hallucinate due to weak schema grounding

This system converts natural language questions into accurate, executable SQL queries using a schema-aware RAG pipeline, also runs python analysis further for better insights if required. Unlike other text-to-SQL systems, it first transforms database schemas into semantic documents, embeds them into a vector store, and retrieves only the most relevant tables for a given query. This context is injected into a structured prompt, ensuring the LLM generates SQL grounded in actual schema constraints. The architecture is modular. it supports multiple entry points (API, CLI, MCP tools, LangChain or LangGraph agents) and can optionally execute queries against a live database. The system also returns structured metadata such as tables used, execution time, and results. By combining semantic retrieval + controlled prompt engineering + execution layer, the platform significantly reduces hallucination and improves reliability for real-world analytics use cases.