Walber Fellipe — AI-focused full-stack engineer

Product showcase

Flows & screens

01 Slack delivery: rendered tables posted as images—heatmaps and KPI columns match the viz config (alignment, currency, percent, conditional colors).

How it's built

Implementation

01

HTML → image pipeline

BigQuery results become styled HTML tables; Puppeteer renders PNGs, uploads to GCS, and Block Kit messages swap in visualization URLs.

02

Scheduling & workers

Cron + timezone aware jobs via BullMQ and Redis, with separate workers for general tasks and Slack execution—including recovery on startup.

03

Templates & safety

Supabase-backed templates and credentials; Zod-validated Fastify routes, Bearer auth, and Prometheus metrics for operability.

Description

This project is a TypeScript microservice that automates Slack delivery from templates stored in Supabase. The typical flow: load templates and credentials, run BigQuery SQL (scoped by company), build HTML tables with conditional formatting, render PNGs with Puppeteer, upload images to Google Cloud Storage, and replace placeholders (such as {{visualization_url}}) in Slack Block Kit payloads before posting. Scheduling uses cron + timezone, asynchronous processing with BullMQ and Redis, a Fastify API with Zod validation, and observability via Prometheus (/metrics) and health checks.

Stack

Node.js 20+, TypeScript, Fastify
BullMQ + Redis (repeatable jobs and on-demand execution)
Supabase (messages, templates, schedules, tokens)
BigQuery, GCS, Slack Web API
Puppeteer for HTML-to-image rendering
Jest for tests

First steps

cd slack-bot-microservice
npm install
cp .env.example .env
# Fill in Redis, Supabase, GCP, GCS, Slack tokens, etc.
npm run dev

API overview

Main routes (most require Bearer authentication):

GET /health — service status
GET /metrics — Prometheus metrics
POST /jobs — create, update, or disable a scheduled job (cron + timezone)
POST /execute-slack-message — enqueue an immediate run (full pipeline)
POST /send-message-simple — send a message by messageId (handy for testing)
POST /test-sql — run BigQuery SQL without posting to Slack
GET /debug-template/:messageId — inspect a template for debugging

Schedule a job (example)

curl -X POST "http://localhost:8080/jobs" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <SUPABASE_SERVICE_ROLE_KEY>" \
  -d '{
    "scheduleId": "YOUR-SCHEDULE-UUID",
    "cron": "0 9 * * 1-5",
    "timezone": "America/New_York",
    "status": "enabled",
    "payload": {}
  }'

Technical highlights

Queues and workers: separate workers for general jobs and Slack message execution; recovery of active jobs on startup.
Slack threads: parent messages with ordered child replies via position.
Visualization: viz_config_json controls alignment, number/currency/percent formats, and conditional color scales on cells.
Deployment: Google Cloud Run with a Puppeteer image (Dockerfile.puppeteer); min-instances so workers are not scaled to zero.

Orca Analytics