0then0
EN

Metrics and usage

OverflowLens

Exact bounds when request outcomes are hidden by cardinality overflow. Work out the possible error rate when OpenTelemetry has merged request attributes into an overflow series.

Release
v0.2.1, Nov 2, 2026
Runs on
Browser app with local file processing; standalone archive available
License
Apache-2.0
Open appDownload standalone archiveView source

Download standalone archive: The app is available in your browser. The published v0.2.1 archive can also be served locally.

Known successes                 900
Known errors                      1
SDK overflow                     99
Total                         1,000

Possible true error rate       0.1% … 10%
Maximum allowed error rate       1%
Conclusion: Cannot determine from these data
Documented result of the bundled synthetic delta example at a 1% threshold; not a measurement from a production system.

What it does

Import a saved OpenTelemetry request counter and compare its possible error rates with your threshold. Delta mode analyzes one collection interval; cumulative mode compares matching Before and After snapshots. Files are parsed locally in a Web Worker; counts and bounds are calculated in your browser. The hosted app downloads its own assets; imported files, threshold values and exported reports are not uploaded.

Who it is for

  • Developers investigating error rates after OpenTelemetry SDK cardinality overflow.

Limitations

  • Accepts uncompressed UTF-8 OTLP JSON monotonic integer Sum counters, up to 5 MiB per file. It is not a live metrics backend or a general SLO evaluator.
  • Cumulative windows require matching complete snapshots with stable attribution, known unchanged starts and no resets or decreasing counts.
  • Provenance assumptions require confirmation. A file cannot prove that a pipeline preserved all counts and overflow points.

Typical uses

Understand a delta interval

Open the bundled example, confirm its assumptions, and see why a 1% limit cannot be decided from bounds of 0.1% to 10%.

Compare cumulative snapshots

Choose Before and After exports of the same continuous counter history. The browser distribution includes a synthetic pair from the SDK lab.

Download

Latest stable release v0.2.1, published Nov 2, 2026. Release notes. Release data checked Nov 2, 2026 from GitHub Releases.

Any platform

Download ZIP archive 96 kB
  • Universal
  • Package
  1. Extract the ZIP and serve its directory over HTTP. Node.js and Go are not required.

Verify a download

SHA256SUMS.txt lists SHA-256 hashes for the release files. A matching hash shows that your file is identical to the one uploaded with the release. It is a checksum, not a signature from the author: it does not prove who built the file.

shasum -a 256 -c --ignore-missing SHA256SUMS.txt

First run

  1. Open the browser app and choose the delta or cumulative example. No SDK or personal files are needed.

  2. For the currently published standalone build, download the ZIP, extract it and run a static HTTP server inside overflowlens-0.2.1:

    python3 -m http.server 8080 --bind 127.0.0.1

  3. Open http://127.0.0.1:8080, choose a mode, confirm the provenance assumptions and set the maximum allowed error percentage. Export JSON to save the explanation.

Platforms and how they were checked

  • Modern web browserDeclared support, no recorded runtime check

    Static application on GitHub Pages; the standalone archive requires an HTTP server, not file://.

Questions and bug reports go to the project’s issue tracker. Report security problems as described in its security policy.