v0.5.0

hosomaki v0.5.0

First public pre-release. Local-only Linux layer intelligence with plain language explanations.

Expect it to be usable, not yet stable.

What it does

Runs system diagnostics locally and explains them using a model served by Ollama (llama3 by default). Every command runs through the same pipeline: collect → sanitise → prompt → validate → repair → render. Raw system data is sanitised before it ever reaches the model.

Commands

• doctor — overall system health check
• explain — explain arbitrary log lines or error output
• why <exit-code> — explain a process exit code
• ports — explain listening ports and what owns them
• timers — explain systemd timers
• crons — explain cron jobs
• mounts — explain mounted filesystems
• updates — explain pending package updates
• firewall — explain active firewall rules
• history — local log of past diagnostics
• audit — system audit
• watch <service> — stream and explain a service live
• status — runtime and status info
• shell-integration — shell hook setup

Notable

• Config via file or env vars (ai.model, ai.endpoint, ai.timeout, output.color, output.language)

Install

.deb / .rpm packages or build from source. Requires Ollama running locally with a model pulled. See the installation docs.
📖 Docs: https://rivernova.github.io/hosomaki/

🔏 Artifacts are checksummed and signed with cosign (keyless, via this workflow's GitHub Actions identity). Verify with:

cosign verify-blob \
  --certificate-identity "https://github.com/rivernova/hosomaki/.github/workflows/release.yml@refs/tags/v0.5.0" \
  --certificate-oidc-issuer "https://token.actions.githubusercontent.com" \
  --bundle checksums.txt.sigstore.json \
  checksums.txt
sha256sum --ignore-missing -c checksums.txt