Skip to main content

Pinche moves to Gemini 2.5 and centralized logs

RestaPro Team
Product & Platform

Invisible but important improvements: the AI assistant is faster, and platform errors no longer disappear into a log nobody checks.

Pinche on Gemini 2.5 Flash​

Pinche now runs on Gemini 2.5 Flash instead of 2.0. You will notice it in:

  • Speed. Responses start appearing about 30% faster.
  • Reasoning. Better when you ask things like "what menu change would have the biggest impact on average ticket?" or "explain this to a new server".
  • Cost. Equal or lower for us, so it does not affect your bill.

The change is transparent. Pinche still answers in Spanish, with the same management tools.

Centralized logs​

Errors from edge functions used to live in six different places: Supabase, Sentry, local nginx logs… Now they all flow into a single #logs channel via n8n:

  • Any error or warning from any serverless function lands in Slack within seconds.
  • If your restaurant shows odd behavior and you email us, we correlate it with the logs from that exact moment.
  • We automatically filter out known noise (expected timeouts) so we do not spam the channel.

It affects you indirectly: when something breaks in production, we see it in minutes instead of hours.

Sentry tunnel via /m​

Some ad blockers block Sentry's domain outright, leaving us blind to frontend errors. Now the frontend sends errors to /m (a same-origin proxy on our nginx) that forwards to Sentry.

Result: we see 100% of the errors happening in the browser, even from users on aggressive uBlock Origin.

ESENCA