Lightweight request signal collection for Ruby applications
Collect signals today. Derive insights tomorrow.

app_pulse is a small, production-safe Ruby gem that tracks HTTP request lifecycle signals using Rack middleware and stores them in simple, file-based formats for offline analysis.

It is not an APM, not an optimizer, and not a dashboard.
It is a signal collector.

• Collect signals, not opinions
• Be lightweight and production-safe
• Zero configuration by default
• No Rails internals in core
• No database writes
• Data first, insights later
• Extensible via future gems

app_pulse is designed to run quietly in production without affecting request flow.

• Tracks HTTP requests via Rack middleware
• Measures request duration
• Captures status and failures
• Stores data in rotating files (CSV / JSON / Text)
• Works with Rails, Rack, Sinatra, and other Rack apps

• No automatic optimization
• No performance advice
• No UI or dashboard
• No DB writes
• No Rails-only behavior

Add to your Gemfile:

gem "app_pulse"

Or for local development:

gem "app_pulse", path: "../app_pulse"

Then run:

bundle install

1. Add middleware

# config/application.rb
config.middleware.use AppPulse::Middleware::Request

That’s it. No initializer is required unless you want custom configuration.

2. (Optional) Configuration

# config/initializers/app_pulse.rb
AppPulse.configure do |config|
  config.output_path = "log/app_pulse"
  config.output_format = :csv
  config.sampling_rate = 1.0
  config.slow_threshold_ms = 500
end

require "app_pulse"

AppPulse.configure do |config|
  config.output_path = "log/app_pulse"
  config.slow_threshold_ms = 500
end

use AppPulse::Middleware::Request

run ->(env) {
  [200, { "Content-Type" => "text/plain" }, ["OK"]]
}

Collected Signals For each request, app_pulse captures:

Output Formats CSV (default)

2026-01-10T14:35:22Z,GET,/health,200,3.2,true,
2026-01-10T14:35:30Z,GET,/error,500,2.1,false,Boom!

JSON

{"timestamp":"2026-01-10T14:35:22Z","method":"GET","path":"/health","status":200,"duration_ms":3.2,"success":true,"error":null}

Text

timestamp=2026-01-10T14:35:22Z | method=GET | path=/health | status=200 | duration_ms=3.2 | success=true | error=

File Rotation

• Files are rotated daily

• One file per format per day

• No file locking

• Append-only writes

Example:

log/app_pulse/
├── 2026-01-10.csv
├── 2026-01-11.csv

Sampling

config.sampling_rate = 0.3

• 1.0 → collect 100% of requests

• 0.3 → collect ~30%

• 0.1 → collect ~10%

Sampling is:

• Stateless

• Thread-safe

• Cheap

• Industry standard

By default, app_pulse collects all sampled requests.

You can optionally configure a slow request threshold to collect only requests that exceed a given duration.

config.slow_threshold_ms = 500

Behavior:

• nil (default) → collect all requests

• 500 → collect only requests taking 500ms or longer

This helps reduce noise in high-traffic applications without changing the meaning of collected signals.

Notes:

• The threshold is applied after sampling

• No errors or insights are inferred

• This is a filtering mechanism, not an optimization

• Errors inside app_pulse never break your app

• Writer failures are swallowed intentionally

• Application exceptions are re-raised

• Observability must never affect availability.

Ruby

• Ruby 2.3+ (tested on 2.3.8 and modern Ruby)

• Designed to avoid modern syntax

Frameworks

• Rails

• Rack

• Sinatra

• Any Rack-compatible framework

Dependencies

• No Rails dependency in core

• No database

• No ActiveSupport

app_pulse follows Semantic Versioning with a deliberately conservative approach.

• v0.x

  • Core behavior is stable

  • Internal structure may evolve

  • New features are additive and opt-in

  • Breaking changes (if any) are documented clearly

• v1.0

  • Core APIs are frozen

  • Extension points are finalized

  • Intended for long-term production use

Version bumps are intentional and documented in the changelog.

• v0.1.x

  • Request lifecycle signal collection

  • Rack middleware

  • CSV / JSON / Text writers

  • File-based storage

  • Configurable sampling

  • Production-safe, fault-tolerant design

• v0.2.x

  • Configurable signal filtering

  • Slow request thresholds

• v0.3.x

  • Aggregation helpers

  • Slow endpoint identification

  • Error frequency summaries

  • Structured data preparation for reports

• v1.0

  • Stable observability core

  • Frozen public APIs

  • Extension-ready architecture

• Future Extensions (separate projects)

  • app_pulse-sql → database query timing

  • app_pulse-jobs → background job tracking

  • app_pulse-exporter → external systems

  • app_pulse-dashboard → UI & reports

• RSpec for unit tests

• Tested with:

  • Rails (modern Ruby)

  • Rack (Ruby 2.3)

• RuboCop compatible

• No CI lock-in

MIT License © Virendra Jadhav

app_pulse is intentionally minimal in its early versions.

Feedback, real-world use cases, and design discussions are welcome. Please open an issue for:

• feature requests
• design questions
• integration ideas

Dashboards and insights are planned as separate projects.

Final Notes

app_pulse is intentionally boring.

That’s a feature.

It gives you raw, trustworthy signals so you can decide what to do with them later — in reports, dashboards, or external systems.