rdmbair15m5-changelog-20260911-0916-procreap-and-xtty-swift6-realization

Realized procreap and xTTY as native Swift 6 applications with strict concurrency and Universal2 fat binaries (arm64 + x86_64), fully verified on canonical hub rdmsm4x:~/dev/apps/, and synchronized to fleet remotes (git.ecs0.net), private GitHub repositories (origin/backup), and rdmbair13m5:~/git-mirror.

Scope & Target Hosts

• Primary host: rdmsm4x (canonical dev hub)
• Operating host: rdmbair15m5 (remote session)
• Affected repositories: ~/dev/apps/procreap, ~/dev/apps/xTTY

Details of Work Performed

1. procreap Native Swift 6 Package Realization:

   • Implemented ProcReapCore utilizing Darwin kinfo_proc, proc_pid_rusage, and sysctl(KERN_PROC_ALL) for zero-subprocess high-speed process table inspection.
   • Built delta CPU rate sampler computing true instantaneous utilization (Δcputime / Δwallclock × 100) avoiding lifetime average distortion.
   • Built zombie grouping by parent PID (ppid) and classification of stopped (SIGSTOP) and uninterruptible wait (U) processes.
   • Implemented safety policy immunizing protected system daemons, active process ancestor chains up to PID 1, and SSH session parents (sshd-session).
   • Built ProcReapCLI executable supporting --nudge (SIGCHLD), --kill-parents, --yes, --json, and Space Blue ANSI reporting.
   • Universal2 fat binary verified via lipo -info (x86_64 arm64).
   • 6/6 unit tests passed in 0.217s under Swift 6 strict concurrency.

2. xTTY Native Swift 6 Package Realization:

   • Implemented XTTyCore with LLMAccount, TokenUsageRecord, QuotaHeadroom, and PrivacyGuard.
   • Enforced strict Privacy Invariant: PrivacyGuard.extractCleanMetrics extracts token accounting fields while immediately discarding all conversational prompts, messages, and code snippets in the same pass.
   • Implemented XTTyDetectors: auto-discovers Claude Code accounts (~/.claude.json), Codex profiles (~/.codex/), Antigravity/Gemini accounts (~/.gemini/), and API key identifiers (~/.secrets/global.env). Accurately reads rate limit tiers (e.g. default_claude_max_20x).
   • Implemented XTTyForecast: BurnRatePredictor computing moving-average tokens/minute, tokens/hour, time-to-exhaustion (TTE), and observation confidence scores.
   • Built XTTyCLI (xtty status, xtty accounts, xtty --json).
   • Universal2 fat binary verified via lipo -info (x86_64 arm64).
   • 3/3 unit tests passed with 0 failures under Swift 6 strict concurrency.

3. Fleet Git & GitHub Synchronization:

   • procreap: initialized bare repo git.ecs0.net:git/apps/procreap.git, created private GitHub repo richhdoty/rdmsm4x-dev-apps-procreap, pushed to fleet and backup.
   • xTTY: committed and pushed to git.ecs0.net:git/apps/xTTY.git, https://github.com/richhdoty/rdmsm4x-apps-xtty.git (origin), and https://github.com/richhdoty/rdmsm4x-dev-apps-xtty.git (backup).
   • Triggered com.eastcoastscience.gitmirror via launchctl kickstart to mirror to rdmbair13m5:~/git-mirror.

Verification Evidence

• procreap tests: 6/6 passed (100% green).
• procreap universal binary: lipo -info .build/out/Products/Release/procreap -> x86_64 arm64.
• procreap live run: accurately categorized 26 zombies across 22 parents, excluding SSH and ancestor sessions, exit code 20.
• xTTY tests: 3/3 passed (100% green).
• xTTY universal binary: lipo -info .build/out/Products/Release/xtty -> x86_64 arm64.
• xTTY live run: discovered [email protected] (tier: default_claude_max_20x), Codex CLI, and local Ollama loopback.

Undo / Rollback

• Each package can be reverted to prior baseline commit via git reset --hard HEAD~1 on its respective repo on rdmsm4x.