Trifle

About Trifle

Trifle is a time-series metrics layer that deliberately refuses to be another observability stack. The pitch is that you already run a database, and product metrics — orders, revenue, signups, churn, background-job success rates, feature adoption by plan — do not justify standing up a separate columnar store and a separate query language to answer. Trifle's libraries write counters and hierarchical values straight into the database you already operate, so a single `Trifle::Stats.track` call with a key, a timestamp and a nested values hash records the order count, the revenue figure, the country breakdown and the acquisition channel in one shot. Three surfaces sit on top of that primitive: Trifle App for dashboards, digests and alerting; Trifle CLI for querying from a terminal or from an AI agent; and Trifle Stats, the open-source instrumentation libraries for Ruby, Elixir and Go. The AI-agent angle is the newest and the most interesting one for this catalogue — Trifle exposes metrics to agents through a local SQLite file and an MCP server, so a coding assistant can read your production KPIs without being handed warehouse credentials. The published case study claims a customer tracking 80 million daily product calculations and 900 million events a day on the design, which is a useful sanity check that the write-into-your-own-database approach scales past hobby volume.

Key Features

Trifle Pros & Cons

Tags