Nixtla BigQuery Forecaster: Cloud Functions on 200 Million Rows

Sat, 29 Nov 2025 10:00:00 -0600

Plugin #2. The Baseline Lab proved statsforecast works on curated datasets. The BigQuery Forecaster proves it works on real data at real scale.

The Plugin

A Cloud Function (Python 3.12) that queries BigQuery public datasets, runs AutoETS and AutoTheta forecasts, and returns structured results. The target dataset: Chicago taxi trips — 200 million+ rows of timestamped ride data with fare amounts, trip durations, and pickup locations.

2,691 lines across 16 files. That’s a complete plugin: source code, tests, deployment config, CI workflow, and documentation. The file count breaks down as: 4 Python source files, 3 test files, 2 deployment configs (Cloud Functions + GitHub Actions), 1 requirements file, and 6 documentation files including the README, architecture doc, and usage examples.

HustleStats Firebase Migration: 5-Agent A2A Architecture on Day One

Mon, 10 Nov 2025 10:00:00 -0600

Four commits to kick off the HustleStats Firebase migration. The previous stack – Next.js on Cloud Run with Prisma and PostgreSQL – worked but cost $72/month for a youth sports app with minimal traffic. Firebase brings that down to an estimated $20/month. The tradeoff is a complete backend rewrite.

The 5-Agent A2A Architecture

HustleStats is a youth soccer statistics tracker. Parents log games, coaches review performance, kids see their highlights. The new backend uses Google’s Agent-to-Agent (A2A) protocol with five agents:

Cloud-Functions on Start AI Tools - Presented by Intent Solutions

Nixtla BigQuery Forecaster: Cloud Functions on 200 Million Rows

The Plugin

HustleStats Firebase Migration: 5-Agent A2A Architecture on Day One

The 5-Agent A2A Architecture