Skip to content

Reporting Issues

Found a path where the model disagrees with your on-air experience? That's exactly the feedback we need.

The Easy Way: ionis-validate report

One command generates a complete, structured report you can paste directly into a GitHub Issue:

# System info + full test suite results
ionis-validate report

# Include your custom path tests
ionis-validate report --custom my_paths.json

# System info only (skip test suite)
ionis-validate report --skip-tests

Progress prints to stderr. The report prints to stdout as markdown — copy it, open an issue, paste it in.

# Pipe directly to clipboard (Linux)
ionis-validate report | xclip -selection clipboard

# Save to file
ionis-validate report --custom my_paths.json > report.md

Where to Report

File an issue on the ionis-training repository. Structured templates guide you through each report type:

https://github.com/IONIS-AI/ionis-training/issues/new/choose

Three templates:

Template When to Use
Test Suite Failure ionis-validate test reports a failure
Prediction Disagreement Model prediction contradicts your on-air experience
Custom Path Results Sharing results from ionis-validate custom

Each template has structured fields — fill in what you can, paste the ionis-validate report output in the "Full Report" field, and submit.

What to Include

Test Suite Failures

If ionis-validate test reports failures:

  1. Run ionis-validate report and paste the full output
  2. Or manually include: ionis-validate info output, failing test ID, expected vs actual

Prediction Disagreements

If ionis-validate predict contradicts your experience:

  1. The exact command you ran (all arguments)
  2. What you expected and why (e.g., "I work this path on FT8 every evening")
  3. Approximate date and UTC time of your observation
  4. Solar conditions if you know them (SFI, Kp)
  5. Station details if relevant (antenna, power class)

Custom Path Results

If your JSON file with expect_open assertions has failures:

  1. Run ionis-validate report --custom your_file.json
  2. The report includes your JSON, the output, and system info — all in one paste

What Happens Next

Every beta test result gets reviewed. Reports are tagged by model version in GitHub Issues, so feedback flows directly into the next version's development cycle.

Paths where the model consistently disagrees with experienced operators become training data for the next version. Your custom path JSON files are particularly valuable — they represent real-world ground truth that no public dataset captures.