How y1ssh/s-agents-practice works
Project Type and Purpose
• This is a prompt-driven “AI Agents Cookbook” repository whose primary focus is on self-contained AI agents.
• The provided “Sql Agent” folder is one agent implementation: it accepts a CSV upload, turns natural-language questions into SQL via an LLM, executes the SQL, and returns results (including visualizations).
File Structure
• Top level:
• README.md – main project overview (“Agents Cookbook”)
• “Sql Agent” directory – one agent’s code and recipe
• Sql Agent folder:
• PROMPT.md – the full LLM prompt (the “recipe”) used to generate the agent
• README.md – agent-specific setup & usage
• requirements.txt – Python dependencies (LLM SDK, pandas, SQL engine, web framework)
• src/
– app.py – web or CLI entry point (API server)
– agent.py – orchestration logic (builds prompt, calls LLM, parses SQL)
– config.py – API keys, model settings, execution parameters
README for Additional Insights
• Agents are run locally by navigating into the agent subfolder.
• Each agent follows the same pattern: a single prompt, a small codebase, and a standard interface.
• The SQL Agent specifically uses an LLM to translate NL → SQL, runs queries against an in-memory or file-backed database (e.g., sqlite/pandas), and returns text + charts.