ShopAgent -- Semana AI Data Engineer 2026

Build a multi-agent AI system that queries structured and semantic e-commerce data -- live, in 4 nights.

What is ShopAgent?

ShopAgent is an autonomous agent crew built on real e-commerce data. It answers business questions by routing to the right data store: SQL for exact numbers, vectors for customer sentiment. Days 1-3 run 100% locally with Docker. Day 4 migrates the same architecture to the cloud.

Central question: O que eu consigo fazer agora que nao conseguia antes?

Architecture

+------------------+     +------------------+     +------------------+
|  DATA GENERATION |     |   AI / LLM       |     |   INTERFACE      |
|  ShadowTraffic   |     |   Claude         |     |   Chainlit       |
+--------+---------+     |   LlamaIndex     |     +--------+---------+
         |               |   LangChain      |              |
         v               |   CrewAI         |              v
+------------------+     +--------+---------+     +------------------+
|  STORAGE         |              |               |   QUALITY        |
|  Postgres        |              v               |   DeepEval       |
|  (The Ledger)    |     +------------------+     |   LangFuse       |
|  Qdrant          |<--->|   MCP Protocol   |     +------------------+
|  (The Memory)    |     +------------------+
+------------------+

The Ledger (Postgres): Exact data -- revenue, counts, averages, JOINs

The Memory (Qdrant): Meaning -- complaints, sentiment, review themes via RAG

Quickstart

Prerequisites: Docker, an Anthropic API key, and a ShadowTraffic license (free trial at https://shadowtraffic.io).

cd gen
cp .env.example .env
cp license.env.example license.env
# Set ANTHROPIC_API_KEY in .env
# Set your ShadowTraffic license fields in license.env
# Get a free trial at https://shadowtraffic.io
docker compose up

Services started: Postgres on 5432, Qdrant on 6333, ShadowTraffic (data generator).

Stack by Day

Day	Theme	Stack
1 Mon	INGERIR	ShadowTraffic, Pydantic, Claude Code, Docker
2 Tue	CONTEXTUALIZAR	LlamaIndex, Qdrant, Postgres, MCP
3 Wed	AGENTE	LangChain, Chainlit, AgentSpec
4 Thu	MULTI-AGENT	CrewAI, DeepEval, LangFuse, Cloud

Data Model

Entity	Store	Fields
customers	Postgres	customer_id, name, email, city, state, segment
products	Postgres	product_id, name, category, price, brand
orders	Postgres	order_id, customer_id (FK), product_id (FK), qty, total, status, payment, created_at
reviews	JSONL -> Qdrant	review_id, order_id (FK), rating, comment, sentiment

3-Agent Crew (Day 4)

Agent	Role	Store
AnalystAgent	SQL data analyst	The Ledger (Postgres)
ResearchAgent	Customer experience researcher	The Memory (Qdrant)
ReporterAgent	Executive report writer	Both via context

Project Structure

gen/                    # Docker infrastructure + data generation
  docker-compose.yml    # Postgres + Qdrant + ShadowTraffic
  shadowtraffic.json    # E-commerce data generators
  init.sql              # Postgres schema
  .env.example          # Environment template
  license.env.example   # ShadowTraffic license template
  data/reviews/         # Pre-generated review data for RAG
docs/                   # Curriculum spec and 4-day agenda
prompts/                # Sequenced live-coding prompts per day
  d1-ingest/            # Day 1: ShadowTraffic + Pydantic (11 prompts)
src/                    # Python requirements per day
presentation/           # HTML slide decks
  d1-ingest.html        # Day 1 slides (143 slides)
.claude/kb/             # 18 knowledge base domains
.claude/agents/         # SubAgents (ai-ml, code-quality, communication, domain, exploration)

---

AIDE Brasil | Formacao AI Data Engineer 2026 | Luan Moreno | April 13-16, 2026