Genomi

Genomi 0.1.0 is the first public release. These notes focus on release-specific
changes and upgrade-relevant behavior; see README.md for the broader product
overview, source support, privacy model, and setup flow.

Highlights

• First public package metadata and install/update path for genomi.
• Local decode dashboard generation is available for broad first-pass genome
  review.
• Active Genome Index parsing now supports variant-ready gVCF imports with
  background reference-block completion.
• .genome/1.0 bundle intake is supported alongside the source formats
  described in the README.
• Pharmacogenomic and ClinVar review paths now report clearer evidence state,
  missing-source state, and sample-specific context.
• Install/update runs are idempotent, so re-running genomi install or
  genomi update is the supported maintenance path.

Decode Dashboard

• Dashboard panels now distinguish ready, empty, blocked, and unavailable
  states.
• Missing libraries, unavailable sources, or insufficient overlap are shown as
  limitations instead of negative findings.
• Pharmacogenomic findings are ordered by finding severity.
• Journal content is no longer included in the dashboard; the dashboard focuses
  on current evidence panels.
• Dashboard output is staged for local browser viewing.

Pharmacogenomics And ClinVar

• Medication-response review separates guideline evidence, label evidence,
  association evidence, sample-specific evidence, and missing evidence more
  clearly.
• PharmCAT workflows handle preflight checks, generated artifacts, imported
  results, calls-only TSV files, and VCF header edge cases more reliably.
• ClinVar matching preserves the connection between public assertions and the
  observed allele evidence from the local genome index.
• Missing optional evidence libraries are reported as blocked evidence scope,
  not as absence of findings.

Genome Indexing

• Source detection is content-based and can handle compressed or archived inputs.
• gVCF parsing becomes usable for variant interpretation before reference-block
  processing is fully complete.
• Consumer-array no-calls and downstream genotype matching have clearer
  behavior.
• VCF export for downstream pharmacogenomic workflows now normalizes metadata
  and invalidates stale export caches.

Install And Update

genomi install and genomi update are the same update path. Re-running either
command is safe and will only refresh what needs refreshing unless a forced
download is requested.

The updater can:

• Refresh the Genomi runtime.
• Install or update selected public reference libraries.
• Repair host-agent skill links.
• Refresh public retrieval indexes.
• Keep older local Active Genome Index records aligned with the current runtime.

Genomi now also follows modern data-directory defaults when available, while
still supporting GENOMI_HOME for explicit installs.

Known Limits

• Genomi is still experimental, and the tool surface may continue to change.
• Some evidence areas require optional public libraries or external tools before
  they can answer.
• External public sources may be temporarily unavailable.
• Missing data, missing libraries, or low overlap are limitations, not negative
  evidence.
• Polygenic, ancestry, nutrigenomic, and common-trait findings should be
  interpreted qualitatively unless a cited source provides a validated number.