SQLGenie

Convert natural language questions into SQL using Amazon Bedrock (Claude 3 Haiku). Includes a safety layer that blocks all destructive operations and returns only SELECT statements.

What Problem This Solves

Writing SQL is a barrier to data access for non-technical users. SQLGenie removes that barrier: type a question in plain English, get a validated, safe SQL query back in under a second, with an explanation of what it does and why.

Safety-first: Only SELECT statements are returned. Any query containing INSERT, UPDATE, DELETE, DROP, TRUNCATE, or ALTER is automatically rejected before it reaches the user.

Demo

Input: "Show me the top 5 customers by total order amount in the last 90 days"

SQL:
  SELECT c.customer_id, c.name, SUM(o.total) as total_spend
  FROM customers c
  JOIN orders o ON c.customer_id = o.customer_id
  WHERE o.created_at >= NOW() - INTERVAL '90 days'
  GROUP BY c.customer_id, c.name
  ORDER BY total_spend DESC
  LIMIT 5;

Explanation: Joins customers with orders, filters to last 90 days,
             groups by customer, sums totals, returns top 5 by spend.

Safety check: PASS - SELECT only, no destructive operations.

Architecture

User question + schema selection
        |
        v
  FastAPI /generate endpoint
        |
        v
  Prompt builder (schema + question -> structured prompt)
        |
        v
  Amazon Bedrock (Claude 3 Haiku)
        |
        v
  SQL extractor (parse SQL from LLM response)
        |
        v
  Safety validator (SELECT-only enforcement)
        |
        |-- PASS -> return SQL + explanation
        |-- FAIL -> return 400 + reason

Key design decisions:

• Schema injected into prompt so Claude understands your table structure
• Safety check runs on extracted SQL, not raw LLM output (avoids false positives)
• Three built-in schemas let you test without a real database
• Stateless API: each request is fully self-contained

Tech Stack

Component	Technology
LLM	Amazon Bedrock (Claude 3 Haiku)
API	FastAPI + uvicorn
Frontend	Streamlit
Containerization	Docker

Quick Start

git clone https://github.com/Rajeshdevandla/sql-genie.git
cd sql-genie
cp .env.example .env
# Fill in your AWS credentials
pip install -r requirements.txt
uvicorn api.main:app --reload

API docs auto-generated at: http://localhost:8000/docs

Required environment variables:

Variable	Description
AWS_REGION	AWS region (e.g. us-east-1)
AWS_ACCESS_KEY_ID	IAM user access key
AWS_SECRET_ACCESS_KEY	IAM user secret key

Run with Docker

docker build -t sql-genie .
docker run -p 8000:8000 --env-file .env sql-genie

API Reference

Method	Path	Description
GET	/health	Health check
GET	/schemas	List available sample schemas
POST	/generate	Generate SQL from a question

Generate SQL:

curl -X POST http://localhost:8000/generate \
  -H "Content-Type: application/json" \
  -d '{"question": "Show me top 5 customers by total order amount", "schema_name": "ecommerce"}'

Response:

{
  "sql": "SELECT customer_id, name, SUM(total) FROM orders GROUP BY 1,2 ORDER BY 3 DESC LIMIT 5;",
  "explanation": "Groups orders by customer, sums totals, returns top 5 by spend.",
  "safety_check": "PASS",
  "schema_used": "ecommerce"
}

Rejected query:

{
  "detail": "Safety check failed: query contains prohibited operation DROP."
}

Built-in Sample Schemas

• ecommerce - customers, orders, products, order_items
• hr - employees, departments, projects
• analytics - events, sessions, users

Project Structure

sql-genie/
|-- config.py              # env var config and validation
|-- core/
|   |-- sql_generator.py   # NL-to-SQL pipeline + safety validation
|-- api/
|   |-- main.py            # FastAPI routes
|-- frontend/
|   |-- app.py             # Streamlit UI
|-- schemas/               # built-in sample schemas
|-- requirements.txt
|-- .env.example
|-- Dockerfile

What I'd Build Next

• Real DB connection - execute validated query against a live database and return actual results
• Schema auto-detection - infer schema from a live DB connection instead of manual input
• SQL parser safety - replace keyword matching with sqlparse for more accurate validation
• Query history - save past queries with ratings for reuse and refinement
• Live hosted demo - deploy to Hugging Face Spaces

Related Projects

• AskDocs AI - PDF RAG chatbot using Amazon Bedrock and FAISS
• AgentFlow - Multi-agent orchestration with Constitutional AI safety layer
• SupportGPT - AI customer support with sentiment-aware escalation

---

Built by Rajesh Kumar - Full Stack Java & AI Developer | Chicago, IL# SQLGenie

Convert natural language questions into SQL using Amazon Bedrock (Claude 3 Haiku). Includes a safety layer that blocks all destructive operations and returns only SELECT statements.

What Problem This Solves

Writing SQL is a barrier to data access for non-technical users. SQLGenie removes that barrier: type a question in plain English, get a validated, safe SQL query back in under a second, with an explanation of what it does and why.

Safety-first: Only SELECT statements are returned. Any query containing INSERT, UPDATE, DELETE, DROP, TRUNCATE, or ALTER is automatically rejected before it reaches the user.

Demo

Input: "Show me the top 5 customers by total order amount in the last 90 days"

SQL:
  SELECT c.customer_id, c.name, SUM(o.total) as total_spend
  FROM customers c
  JOIN orders o ON c.customer_id = o.customer_id
  WHERE o.created_at >= NOW() - INTERVAL '90 days'
  GROUP BY c.customer_id, c.name
  ORDER BY total_spend DESC
  LIMIT 5;

Explanation: Joins customers with orders, filters to last 90 days,
             groups by customer, sums totals, returns top 5 by spend.

Safety check: PASS - SELECT only, no destructive operations.

Architecture

User question + schema selection
        |
        v
  FastAPI /generate endpoint
        |
        v
  Prompt builder (schema + question -> structured prompt)
        |
        v
  Amazon Bedrock (Claude 3 Haiku)
        |
        v
  SQL extractor (parse SQL from LLM response)
        |
        v
  Safety validator (SELECT-only enforcement)
        |
        |-- PASS -> return SQL + explanation
        |-- FAIL -> return 400 + reason

Key design decisions:

• Schema injected into prompt so Claude understands your table structure
• Safety check runs on extracted SQL, not raw LLM output (avoids false positives)
• Three built-in schemas let you test without a real database
• Stateless API: each request is fully self-contained

Tech Stack

Component	Technology
LLM	Amazon Bedrock (Claude 3 Haiku)
API	FastAPI + uvicorn
Frontend	Streamlit
Containerization	Docker

Quick Start

git clone https://github.com/Rajeshdevandla/sql-genie.git
cd sql-genie
cp .env.example .env
# Fill in your AWS credentials
pip install -r requirements.txt
uvicorn api.main:app --reload

API docs auto-generated at: http://localhost:8000/docs

Required environment variables:

Variable	Description
AWS_REGION	AWS region (e.g. us-east-1)
AWS_ACCESS_KEY_ID	IAM user access key
AWS_SECRET_ACCESS_KEY	IAM user secret key

Run with Docker

docker build -t sql-genie .
docker run -p 8000:8000 --env-file .env sql-genie

API Reference

Method	Path	Description
GET	/health	Health check
GET	/schemas	List available sample schemas
POST	/generate	Generate SQL from a question

Generate SQL:

curl -X POST http://localhost:8000/generate \
  -H "Content-Type: application/json" \
  -d '{"question": "Show me top 5 customers by total order amount", "schema_name": "ecommerce"}'

Response:

{
  "sql": "SELECT customer_id, name, SUM(total) FROM orders GROUP BY 1,2 ORDER BY 3 DESC LIMIT 5;",
  "explanation": "Groups orders by customer, sums totals, returns top 5 by spend.",
  "safety_check": "PASS",
  "schema_used": "ecommerce"
}

Rejected query:

{
  "detail": "Safety check failed: query contains prohibited operation DROP."
}

Built-in Sample Schemas

• ecommerce - customers, orders, products, order_items
• hr - employees, departments, projects
• analytics - events, sessions, users

Project Structure

sql-genie/
|-- config.py              # env var config and validation
|-- core/
|   |-- sql_generator.py   # NL-to-SQL pipeline + safety validation
|-- api/
|   |-- main.py            # FastAPI routes
|-- frontend/
|   |-- app.py             # Streamlit UI
|-- schemas/               # built-in sample schemas
|-- requirements.txt
|-- .env.example
|-- Dockerfile

What I'd Build Next

• Real DB connection - execute validated query against a live database and return actual results
• Schema auto-detection - infer schema from a live DB connection instead of manual input
• SQL parser safety - replace keyword matching with sqlparse for more accurate validation
• Query history - save past queries with ratings for reuse and refinement
• Live hosted demo - deploy to Hugging Face Spaces

Related Projects

• AskDocs AI - PDF RAG chatbot using Amazon Bedrock and FAISS
• AgentFlow - Multi-agent orchestration with Constitutional AI safety layer
• SupportGPT - AI customer support with sentiment-aware escalation

---

Built by Rajesh Kumar - Full Stack Java & AI Developer | Chicago, IL