Project: Industrial Safety & Surveillance → SQL Intelligent Assistant (Gemini) through RAG Application
Technologies: Python, MySQL, PyMySQL, LangChain, Gemini-2.5, dotenv, Rich, Pandas, PIL
Duration: Dec 2025 – Present
llm_rag_sql_assistant is a conversational AI assistant that allows users to query MySQL databases using natural language. Leveraging Gemini LLM and RAG (Retrieval-Augmented Generation) principles, it safely converts natural language questions into optimized SQL queries and provides rich output in the terminal.
The assistant supports multi-turn conversations, intelligent SQL generation, automatic error correction, and advanced query capabilities, making it ideal for industrial safety and surveillance applications or any scenario requiring fast insights from databases.
-
Convert natural language queries into safe SQL for MySQL.
-
SQL safety checks to prevent destructive commands (
DROP,TRUNCATE,DELETE,UPDATE). -
Auto-LIMIT for large query results.
-
Support for advanced queries:
- Retrieve last N records or latest data
- COUNT queries and aggregation
- Date range filtering from natural language input
-
LLM-driven error correction for SQL queries on database errors.
-
Rich CLI interface:
- Table visualization using Rich
- CSV export for query results
- Image handling from database blobs, file paths, or URLs
-
Multi-turn conversational support with session memory and follow-up queries.
-
Applied prompt engineering for Gemini LLM to generate structured, syntactically correct SQL.
-
Prompt engineering for structured and syntactically correct SQL output.
-
Impact:
- Reduced manual SQL query effort by 80% for non-technical users.
- Enabled fast insights and reporting from MySQL databases using natural language commands.
- Improved productivity and accuracy by combining AI-driven query generation with safe database practices.
- Clone the repository
git clone git@github.com:Roubish/llm_rag_sql_assistant.git
cd llm_rag_sql_assistant