How rodzanto/agentic-ai-data-transform works
Project type and purpose
• From the file tree and README, this is a full-stack ETL system with an AI-powered natural language interface.
• Purpose: Allow users to upload structured data, issue plain-English transformation requests, and have the system generate and execute Athena SQL via an MCP bridge and Amazon Bedrock (Anthropic Claude).
File structure
• Top level:
• README.md – project overview and architecture description
• config.json – central configuration for S3 buckets, SQL templates, etc.
• claude_code_start.sh – bootstraps Bedrock/Claude context
• requirements.txt – Python dependencies (Streamlit, AWS SDK, MCP client)
• transactions.csv – sample dataset
• function_calling/ directory:
• streamlit_app.py – frontend UI and orchestration logic
• mcp_client_script.py – code that sends generated SQL from the AI agent to the MCP server
• No heavy microservice splits – single UI service, single MCP client, AWS services.
README for additional insights
• AWS services used:
• Amazon Bedrock (Anthropic Claude) as LLM orchestrator
• MCP server as “protocol adapter” for SQL calls
• Amazon Athena for query execution
• Amazon S3 for raw/processed data storage
• Streamlit for UI
• Workflow clearly spelled out in “Architecture Details” and “Workflow” sections.
Detail without over-splitting
• Five major components plus the user and configuration artifacts.