A conversational AI agent for natural language SQL analytics on bank customer data, powered by Gemini (Google Generative AI) and LangGraph's ReAct agent framework.
Here's what the app looks like:
- Ask natural language questions about customer data (e.g., "Top 5 customers by balance", "Average balance by region").
- Agent automatically translates your question into a safe SQL query, runs it, and explains the result.
- Handles grouping, aggregation, filtering, and multi-factor analytics.
- Uses a single tool (
execute_sql) for true agentic reasoning—no manual SQL needed! - Robust to case sensitivity in text columns (thanks to prompt design).
- Data source: CSV (
customers.csv) → SQLite (bank_data.db).
flowchart TD
User["User: Asks Data Question"]
Agent["ReAct Agent (Gemini, LangGraph)"]
Tool["Tool: execute_sql"]
DB[(SQLite DB: bank_data.db)]
Result["Agent: Final Answer (Table + Insights)"]
User -->|Natural Language| Agent
Agent -->|Generates SQL Query| Tool
Tool -->|Executes SQL| DB
DB -->|Results| Tool
Tool -->|Table/Result| Agent
Agent -->|Readable Answer + Table| Result
Result --> User
-
Clone this repo and enter the directory:
git clone https://github.com/nishanttomar21/customer-data-sql-agent.git cd customer-data-sql-agent -
Install dependencies:
pip install -r requirements.txt
-
Prepare your data:
- Place your
customers.csvin the project root (headers must match those described in the script).
- Place your
-
Set your Gemini API key:
- In the script, replace
YOUR_GOOGLE_API_KEYwith your own Google Generative AI API key.
- In the script, replace
-
Run the agent:
python main.py
-
Ask questions:
💬 Ask a data question: Top 5 customers by balance 💬 Ask a data question: What is the average balance by region?
- Show all customers.
- List customers from England.
- What is the average balance by gender, age group, and job classification?
- Which region has the highest number of "White Collar" customers?
- Who joined before 2018?
- Find the date joined for Ruth Campbell from Wales.
- Python 3.8+
- Google Generative AI API Key
customers.csvin the correct schema
- Edit the system prompt for more/less restrictive behavior or change analytics focus.
- Add new columns to
customers.csvand update the prompt’s<database_schema>accordingly.
End-to-end data pipeline for handling large-scale, complex data
flowchart TD
A[Data Sources: ERP, POS, Market, Social] --> B[Ingestion Layer: Batch and Streaming ETL]
B --> C[Cloud Storage or Data Lake: S3, ADLS, GCS]
C --> D[Databricks: Transform and Enrich Data]
D --> E[Snowflake: Cloud Data Warehouse]
E --> F[BI and Data Science: Tableau, Power BI, ML]