[filipeforattini]

Why

Recurring Blacksmith matrices, nightly workloads, fuzzing, and trunk checks made runner spend unbounded relative to normal development activity. This repository is public, so standard GitHub-hosted runners are free.

What changes

• run the required Test Suite, quality, RQL conformance, and Python driver checks on ubuntu-24.04
• run weekly DR, DST, parser-fuzz, and KV benchmark schedules on ubuntu-24.04
• gate feature matrices, driver conformance, containers, fuzz smoke, packaging, Windows, and macOS behind workflow_dispatch with full_ci=true
• make release candidates manual while preserving the stable-release Blacksmith build matrix
• move lightweight orchestration, coverage, docs, Dependabot, and publish jobs to GitHub-hosted runners
• add a manual Blacksmith Testbox workflow for explicit 2/4/8-vCPU measurements
• update workflow contracts

Blacksmith remains available for explicit full-CI, performance, Testbox, and stable-release workloads. The live parser-fuzz and release-candidate workflows remain manually disabled until this rollout lands.

Validation

• actionlint across changed active workflows
• node --test scripts/release_tooling_contract.test.mjs — 15/15 passing
• git diff --check

No release or deployment was invoked.

[filipeforattini]

Live validation on the final head (b5f9c45): the full required CI trunk completed successfully on GitHub-hosted ubuntu-24.04, including the 10,140-test nextest layer, doctests, and persistent multimodel tests. Test Suite finished in 14m49s inside the 20-minute cap. Every Blacksmith-only full_ci job was skipped: https://github.com/reddb-io/reddb/actions/runs/30970339579