Trifle – Time-Series Analytics Without the Infrastructure

TL;DR

TL;DR: Trifle is an open-source time-series analytics library that aggregates nested counters into pre-defined time buckets using the database you already run — Postgres, Redis, MongoDB, MySQL, or SQLite. No InfluxDB, no TimescaleDB, one gem call and you are tracking.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: trifle.io
• Source repository (Ruby): github.com/trifle-io/trifle-stats
• Source repository (Go): github.com/trifle-io/trifle_stats_go
• Source repository (Elixir): github.com/trifle-io/trifle_stats
• CLI: github.com/trifle-io/trifle-cli
• License: MIT (verified via LICENSE files in each repo)
• HN launch thread: news.ycombinator.com/item?id=49007574
• Source last checked: 2026-07-23

What Is Trifle?

Trifle is a time-series analytics library that tracks nested counters and breakdowns without requiring a dedicated time-series database. Instead of ingesting raw events, you call track with aggregated values and Trifle writes them into pre-defined time buckets — one write, multiple rollups.

The core idea, from the HN launch post:

“Aggregate counters into pre-defined time buckets, so a single write increments multiple buckets at once.”

Trifle started in 2015 as a Rails APM project. After two rebuilds over a decade, it now tracks approximately 1 billion events per day at the author’s day job. The library is available in Ruby, Elixir, and Go, with a unified DSL across all three drivers.

Key Features

• No new infrastructure — uses your existing Postgres, Redis, MongoDB, MySQL, or SQLite database
• Nested breakdowns — track revenue by country by channel in a single track call
• Automatic rollup — granularities of 1 minute, 6 hours, 1 day, 1 month handled automatically
• Multi-language — Ruby, Elixir, and Go implementations with the same API surface
• MCP server — AI agent integration via Model Context Protocol for metric payloads and execution tracing
• CLI tool — terminal access to your metrics without a browser
• Agent Skills — npx -y @trifle-io/install brings Trifle best practices to Claude Code, Cursor, Codex, and other coding agents

Supported Databases

| Driver | Status | |---|---| | Postgres | Stable | | Redis | Stable | | MongoDB | Stable | | MySQL | Stable | | SQLite | Stable |

Setup

Ruby / Rails

# Gemfile
gem 'trifle-stats'

# config/initializers/trifle_stats.rb
Trifle::Stats.configure do |config|
  config.driver = Trifle::Stats::Driver::Postgres.new(ActiveRecord::Base.connection)
  config.granularities = ['1h', '1d', '1w', '1mo']
end

# Track an event
Trifle::Stats.track(
  key: 'orders',
  at: Time.now,
  values: {
    count: 1,
    revenue: 49_90,
    revenue_by_country: { us: 49_90 },
    revenue_by_channel: { organic: 49_90 }
  }
)

# Query aggregated values
Trifle::Stats.values(
  key: 'orders',
  from: 1.week.ago,
  to: Time.now,
  granularity: :day
)
#=> { at: [Mon, Tue, Wed, ...], values: [{ "count" => 12, "revenue" => 598_80, ... }, ...] }

Go

package main

import (
    "time"
    stats "github.com/trifle-io/trifle_stats_go"
)

func main() {
    driver := stats.NewPostgresDriver(db)
    client := stats.NewClient(driver, []string{"1h", "1d", "1w", "1mo"})

    client.Track("orders", time.Now(), map[string]interface{}{
        "count":   1,
        "revenue": 49_90,
    })
}

Elixir

defmodule MyApp.Analytics do
  use Trifle.Stats

  def track_order do
    Trifle.Stats.track(
      key: "orders",
      at: DateTime.utc_now(),
      values: %{
        count: 1,
        revenue: 49.90
      }
    )
  end
end

How Nested Breakdowns Work

The nested values hash is the core differentiator. A single call writes to multiple counter dimensions simultaneously:

Trifle::Stats.track(
  key: 'requests::aws::s3_uploads',
  values: {
    count: 1,
    status: { request.response_code => 1 },
    size: payload.bytes,
    duration: { sum: request.duration, count: 1 }
  }
)

This increments the overall count, tracks the distribution of HTTP response codes, logs total bytes uploaded, and records duration metrics — all in one write. Querying the data later lets you slice by any dimension you tracked.

MCP Server for AI Agents

Trifle ships an MCP server that lets AI coding agents interact with your metrics:

npx -y @trifle-io/install

This installs Trifle Skills for supported agents, providing:

• Best practices for metric payload design
• Execution tracing integration
• CLI analytics access

Pricing

The Trifle libraries (Ruby, Elixir, Go) are open-source and free to use. The CLI is also free. Trifle Cloud Pro is available for dashboards and hosted infrastructure:

• Cloud Pro: $19/month (launch special — was $399/month)
• Libraries and CLI: Free (MIT license)
• Self-hosted: Free (run on your own infrastructure)

Comparison to Other Approaches

| Approach | Infrastructure | Nested Counters | Languages | |---|---|---|---| | Trifle | Your existing DB | Yes | Ruby, Elixir, Go | | InfluxDB | Dedicated DB | Limited | Many | | TimescaleDB | Dedicated DB | No | SQL only | | AWS CloudWatch | Managed service | No | API only |

FAQ

Q: Does Trifle require a dedicated time-series database? A: No. Trifle uses your existing Postgres, Redis, MongoDB, MySQL, or SQLite database as its storage layer.

Q: How does Trifle handle high write volumes? A: The author reports that Trifle tracks about 1 billion events per day at their day job using Postgres as the backend.

Q: Which programming languages are supported? A: Ruby (trifle-stats gem), Elixir (trifle_stats), and Go (trifle_stats_go) are all available with a unified DSL.

Q: Is there a managed cloud option? A: Yes, Trifle Cloud Pro is available at $19/month (launch pricing). The libraries and CLI remain open-source and free.

Q: Can AI agents use Trifle? A: Yes. The @trifle-io/install package provides MCP server integration for Claude Code, Cursor, Codex, and other agents that support the Model Context Protocol.

Conclusion

Trifle solves a specific problem well: you want time-series analytics without deploying and maintaining a dedicated time-series database. By aggregating counters into pre-defined buckets at write time, it keeps queries fast even at high event volumes. The nested breakdown support is unusually powerful for a library this lightweight.

If you are already running Postgres, Redis, or MongoDB, adding Trifle takes one gem install and a configuration block. The MIT license and multi-language support make it straightforward to embed in projects without vendor lock-in.

• Trifle project page
• Ruby gem
• Go library
• HN Discussion