Skip to content

Releases: obakeng-develops/mino

Mino v1.0.1

Choose a tag to compare

@obakeng-develops obakeng-develops released this 02 Jul 07:31
Immutable release. Only release title and notes can be modified.
a336389

Post-launch fixes and polish, from a week of running Mino against real incidents.

Incidents

  • Diagnoses no longer hang on "pending." A container that dies without logging — most often an OOM-kill — now gets a diagnosis from its state (exit code / OOM flag) instead of waiting forever for logs that never come.
  • The card matches the decision. When a restart won't fix a failure (OOM, missing config, a URL check), Mino says so and doesn't offer a restart — the heading, the timeline, and the buttons all agree. The approve button is now just "Approve".
  • Collapsible machine groups. The Incidents page groups by machine and collapses by default, with expand/collapse-all, so a large fleet stays scannable. Collapsed rows show count, last-incident time, and a "needs you" marker.
  • Redesigned "happening now" card and skeleton loaders on Incidents and Guardrails.

Now / Fleet

  • The Now page shows the server row even with a single server (name, health, click-through), instead of hiding the breakdown until you have two.

Guardrails

  • Updated to reflect what Mino actually does: it runs restart / stop / start (not restart only), and it won't restart a failure a restart can't fix — it diagnoses and hands off.

Under the hood

  • Dead-code cleanup and a repo-wide tidy.

Mino v1.0.0

Choose a tag to compare

@obakeng-develops obakeng-develops released this 30 Jun 14:42

First open-source release of Mino.

Mino keeps your Docker containers running. It watches them, works out why one failed, and restarts it within the limits you set — or stops and asks first, depending on the autonomy you choose. You and your team see the same fleet, with incidents and the agent's reasoning as they happen.

Highlights

  • Watch servers with the agent, or watch any endpoint with a URL health check.
  • Automatic diagnosis and recovery, with guardrails the agent obeys on every restart.
  • Auto-fix or ask-first autonomy, per fleet and per server.
  • First-run onboarding that walks a new owner to their first watched service.
  • Incidents grouped by machine; a fleet view with uptime at a glance.

Running it

Runs as a single always-on instance plus a Postgres database. See the README "Requirements" section and the deploy guides (Docker Compose, Fly, Kamal). An LLM key is optional — monitoring and restarts work without one.