The private playbook repository grew through repeated regeneration. A directory listing is not an operating procedure, and publishing the source tree unchanged would expose material whose identity and reuse terms were not reviewed.
- Baseline e450849 contains 922 tracked files and 48 media assets.
- History shows consecutive "Regenerate all playbooks" commits.
- The tree separates access, framework setup, multi-system networking, serving, quantization, and application workflows.
- This public adaptation records scope and evidence boundaries while omitting live identifiers and non-original assets not cleared for redistribution.
One generic quick start could represent the collection. The source disproves that: each domain has different prerequisites and acceptance checks.
Count tracked paths and media in the frozen source; inspect representative workflow pages; validate public links and JSON; scan addresses, paths, identities, and secrets; review the Mermaid boundary for actionable topology.
Accept this record when those checks pass. Do not interpret acceptance as proof that a live Spark workflow ran.
The result is an evidence-led index, not a replacement for vendor documentation. Hardware behavior, versions, and model availability still require a fresh operator preflight.