Home

Flakemetry

OpenTelemetry-native test intelligence platform. Treat every test run as a trace, not a report.

Flakemetry (flaky + telemetry) is an open-source platform for test observability, explainable flaky-test detection, and AI-assisted root-cause analysis. Instead of parsing JUnit XML into a static HTML report, Flakemetry models every test execution as an OpenTelemetry span — which makes historical analytics, flaky scoring, and AI root-cause natural extensions of the telemetry rather than bolted-on hacks.

The one-paragraph pitch

Test reporters answer "what happened in this run?" Flakemetry answers "is this test trustworthy, why is it failing, and is it getting worse over time?" — across every run, every branch, and every refactor. It ships as an OSS platform you can run with a single docker compose up, integrates into CI through a GitHub Action, and layers AI root-cause analysis on top of a transparent, explainable flaky-detection model.

Why it exists

Three systemic failures in test tooling:

1. Tests are treated as report artifacts, not telemetry signals. No native time-series, no trace context, no correlation with application telemetry.
2. Flaky detection is primitive. Most teams "detect" flakes by eyeballing retries > 0. No stable test identity across refactors, no statistical model, no auto-quarantine.
3. Root-cause analysis is manual archaeology. 20–40 minutes per incident digging through logs, stack traces, screenshots, and git blame.

Read the full reasoning in Product Vision.

Start here

I want to…	Go to
Understand the product & positioning	Product Vision
See the system design	Architecture
Understand the database	Data Model
Understand the flaky algorithm	Flaky Scoring
Understand test identity	Test Identity Engine
Understand the AI layer	AI RCA Architecture
See the OTel span model	OTel Test Conventions
Understand scaling	Ingestion and Scaling
See the plan	Roadmap
Understand the business model	Monetization and OSS Model
Look up a term	Glossary · FAQ

Project status

Early-stage, actively designed in the open. The full backlog lives in Issues and on the public Roadmap board, organized by milestone M0 → M6.

Tech stack

TypeScript · Playwright · Node.js · PostgreSQL (Prisma) · React / Next.js · Docker · GitHub Actions · OpenTelemetry