TSmithCode AI Workstation Profiler

Repeatable workstation-readiness evidence for AI, local-model, container, and engineering-software development.

This repository turns a Mac or developer workstation into an auditable proof surface: hardware and tool inventory, practical benchmarks, optional workflow checks, install recommendations, sanitization controls, and public-safe reports in Markdown and JSON.

Artifact class: Operational proof.

Evaluator question

Can this workstation support the intended development workflow, and can its readiness be measured, repeated, and shared without exposing private machine data?

Decision this proves

Run this repository to decide whether the machine is ready for the next development slice, which missing tools require approval, and which results are safe to use as technical evidence.

Best for

• AI and local-model developers evaluating a new workstation.
• Technical leaders validating container, data, media, or CAD/BIM/GIS-adjacent readiness.
• Recruiters and hiring managers evaluating environment inspection, automation, benchmarking, and evidence discipline.
• Consulting evaluators who want to see how TSmithCode turns setup work into a repeatable operating system.

Run locally

git clone https://github.com/tsmithcode/tsmithcode-ai-workstation-profiler.git
cd tsmithcode-ai-workstation-profiler
./scripts/check_release_safety.sh
./scripts/run_live_demo.sh

The live demo profiles the machine, runs practical benchmarks, generates install recommendations, builds the public report, and saves a reviewable transcript.

Expected evidence

Generated artifacts are written under results/ and ignored by Git until deliberately reviewed:

profile_*.md
profile_*.json
benchmarks_*.md
benchmarks_*.json
install_recommendations_*.md
install_recommendations_*.json
public_workstation_report_*.md
live_demo_transcript_*.txt

The evidence packet supports human review, machine-to-machine comparison, video recording, and downstream technical evaluation.

What the profiler measures

Environment inventory

• macOS version and model identifier;
• CPU, physical and logical cores, and memory;
• disk capacity and free space;
• relevant applications and CLI tools;
• installed local Ollama models when available.

Practical benchmarks

• single-core and multi-core SHA-256 throughput;
• Python memory-copy bandwidth;
• local sequential disk write/read sanity checks.

Optional workflow checks

When the tools are installed, the profiler can check Docker execution, QGIS processing, GDAL, PDAL, Blender CLI, and an Ollama model response. Missing optional tools are reported as honest readiness signals rather than converted into false success.

Recommendation boundary

./scripts/recommend_installs.py

The recommendation pass does not install software. It separates observation from approval so environment changes remain visible, reversible, and owner-controlled.

Command reference

./scripts/profile_machine.py
./scripts/benchmark_machine.py
./scripts/recommend_installs.py
./scripts/generate_public_report.py
./scripts/run_live_demo.sh
./scripts/check_release_safety.sh
./scripts/build_release_zip.sh

Equivalent Make targets are available:

make profile
make benchmark
make recommend
make report
make demo
make zip

Quality and privacy controls

The release-safety gate checks executable scripts, ignored result folders, local-path leakage, release ZIP construction, and release ZIP validation.

The profiler does not intentionally collect credentials, browser history, email, project files, SSH keys, API keys, or environment-variable dumps. Public reports sanitize home paths, hostnames, and user-specific /Users/<name> paths.

• Benchmark methodology
• Privacy and sanitization
• Quality gate

Recording and communication assets

The same proof system supports GitHub review, a controlled screen-recording workflow, and TSmithCode.ai authority content:

• Director's treatment
• Filming runbook
• One-person crew briefing
• Zero-drift action checklist
• YouTube package

Use ./scripts/start_recording_session.sh to open the recording route and run the safety gate from one entry point.

Proof boundary

This is an operational workstation-readiness proof, not a hardware certification, procurement benchmark, production-capacity guarantee, or customer-result claim. Results are machine-specific, time-specific, and dependent on the installed software and local conditions at the time of execution.

What to send

For a technical review, send only:

• the repository link;
• the newest public workstation report;
• the matching benchmark report;
• the release-safety result;
• one sentence naming the workflow the machine must support.

Do not send raw private machine inventories, credentials, unrestricted screenshots, project files, or unreviewed transcripts.

Related TSmithCode proof system

• TSmithCode software proof kits
• Software discovery diagnostic
• M5 Max video command center
• TSmithCode public proof standard

For CAD, Autodesk, SolidWorks, BIM, Vault, or PDM/PLM implementation work, use CAD Guardian services.

Repository structure

.
├── docs/        Methodology, privacy, quality, and recording guidance
├── examples/    Public-safe examples
├── results/     Local generated outputs, ignored except for .gitkeep
├── scripts/     Profiling, benchmark, reporting, recommendation, and release tools
├── Makefile     Command aliases
├── README.md    Evaluator entry point
└── LICENSE      MIT license

License

MIT. Use it, fork it, and benchmark your own workstation.