What problem does it solve? Building AI-powered Python applications involves juggling model calls, tool loops, streaming, session state, and debugging without visibility. This Skill provides structured guidance for developing with the Genkit Python SDK, covering everything from first setup to multi-turn agents with human approval, plus trace-based debugging workflows. ## Core Features & Use Cases - Agent Development: Define multi-turn agents with tools, typed session state, human-in-the-loop approvals, branching, background tasks, and artifacts using ai.define_agent. - Flows & Streaming: Create single-shot flows with structured output, streaming via ctx.send_chunk, and serve them over FastAPI with SSE. - Dev Workflow & Debugging: Run apps under genkit start to capture traces, inspect model I/O with genkit trace:get, and resolve common errors like wrong imports or missing model prefixes. - Use Case: You want to build a support chatbot in Python that remembers conversation history, asks for approval before executing sensitive tools, and can be served over HTTP. This Skill walks you through defining the agent, attaching a session store, adding ToolApproval middleware, and verifying behavior with traces. ## Quick Start Ask the assistant to create a Genkit Python app that defines an agent with a tool and runs it under genkit start with trace inspection.
