Velune Trace v0.5.0

Velune Trace v0.5.0

Release Focus

v0.5.0 focuses on fresh-clone reproducibility and documentation
accuracy rather than new evidence-extraction features.

Included

• Local MCAP inspection, chunk/read access, and topic timing profiling
• Windowed timing-irregularity ranking and evidence-window extraction
• Raw-window compare / compare-all diagnostics between two MCAP
  windows
• Versioned Core Report Bundle generation via validation-report
• Deterministic sample MCAP generation for local evaluation

Fixed

• pytest no longer depends on an undocumented pre-existing
  examples/sample.mcap. A test-session bootstrap
  (tests/conftest.py) now generates the deterministic sample MCAP
  automatically on a fresh clone, before any test that needs it runs.
• docs/architecture.md contained malformed, stale content that did
  not match the current implementation. Replaced with an accurate
  description of the current MCAP evidence pipeline and the legacy
  JSONL/SQLite indexing pipeline.

Removed

• Two unused, undocumented CLI modules (verify_dataset.py,
  dataset_report.py) that were never wired into the velune
  command dispatcher, had no tests, and had no documentation. Their
  sole dependency, PyYAML, was also removed from
  requirements.txt / requirements-lock.txt. The public command
  set is unchanged.

Release validation

• Full automated test suite passes on a genuinely fresh clone (no
  pre-existing sample MCAP, no pre-existing virtual environment)
• Sample MCAP creation passes
• Local Validation Report generation passes
• All six required Core Report artifacts are present

Product boundary

Velune Trace does not automatically determine:

• root cause
• fault or liability
• safety or severity
• normality or superiority
• regression or improvement

Find the events. Engineers find the cause.