A practical walkthrough of building an LLM-powered meeting scheduler in Go using function/tool calling. The agent queries a local meeting database via a `meetings` tool, passes results back to the LLM in a loop until no more tool calls are needed, and returns available 45-minute time slots. Uses Ardan Labs' Kronk model server (OpenAI-compatible API) with the Ministral model. Covers system prompt design, tool definition schema, argument parsing, and the multi-round-trip conversation loop pattern. Full code available on GitHub in ~160 lines.
Table of contents
IntroductionHow Does It Work?Setting UpApplication OverviewSystem PromptTool DefinitionTool CallingFinding a MeetingMainSummary20K Impressions