AI Agents for Beginners, Part 2: The agent picks between three database tools, uses each result to choose its next call, and stops at an eight-turn circuit-breaker if it gets stuck.
This tutorial expands a shop database to include two customers and four orders, requiring the agent to find Sarah Jones's largest order without being given IDs. The model must dynamically choose from narrow, parameterized SQL tools—such as finding a customer ID, listing orders, and calculating totals—rather than using a broad, unrestricted run_sql() function. This design prevents infinite loops and limits the model's potential for error.
To prevent a single bad decision from triggering hundreds of model calls, the Python control flow implements a hard cap of eight model turns. The agent loop queries the LLM (qwen3.5:4b via Ollama) after each tool result, continuing only if the response contains further tool calls. Once the loop reaches the limit or the model returns a final answer, the process terminates safely.
Source: Towards Data Science · Summarized by HeadlinesBriefing