Focus: AI & LLM orchestration patterns

Level: Intermediate

Prerequisites: Basic Python knowledge, familiarity with Airflow concepts helpful but not required

Environment: Browser-based (Astro IDE), no local setup or Docker required

mission breakdown

Setting the stage:

We'll kick off with an introduction to AstroTrips, Airflow as a unified control plane for data & AI, & the decorator-based approach to LLM orchestration.

Hands-on Workshop Stages:

Setup: Connect to the Astro IDE, start your Airflow environment, & explore the Support Portal plugin.

Analyze reviews: Build an LLM pipeline that classifies & summarizes customer reviews, including image analysis for reviews with attachments.

Route reviews: Use AI-powered branching to assign reviews to the right team, with dynamic task mapping over the full review queue.

Embed reviews: Generate vector embeddings & store them for similarity search, with asset-triggered scheduling.

Respond to reviews: Build an agent with database tools that drafts customer responses, then pause for human approval before persisting the result.