GO3D CLOUD · LOCAL AI · PARALLEL PROCESSING
Start with the local mesh-inspection worker. Reports, setup guides and future plans are below.
Local inspection available · guided installer coming soon
Read an STL or OBJ file and create a separate JSON report with its checksum, geometry counts and coordinate bounds. Local inspection needs no sign-in or network connection. This is not CAD conversion, mesh repair, GPU processing or an accuracy certificate.
Extract the ZIP. Open a terminal in the folder containing dragonfire_worker and follow the included README. Python 3.10 or later is required; there are no third-party Python dependencies.
python -m dragonfire_worker inspect model.stl --output model-inspection.json
Use your own input filename and a new report filename. The worker does not overwrite the original. Geometry units remain unspecified; watertightness and engineering accuracy are not checked.
Connected jobs are an operator-only pilot requiring pairing and separate authorization. Downloading the ZIP does not connect your computer or provide hosted processing. The ZIP includes source, a README and tests; a commercial redistribution license is not included.
Hardware scope, software compatibility and a quote worksheet. The master architecture stays with GO3D.
Open current brief ↗Read the PDF ↗The system, the assumptions, the industry applications and the work still needed to validate performance.
A light-background graphic for reading, sharing and presentations. Includes the planning assumptions.
The matching comparison in charcoal, silver and cyan. The same numbers and qualifications.
One link to this complete collection.
Select and copy this address.
OPERATOR EFFORT ≠ ELAPSED TIME
This example assigns the same accepted processing package to three workflows. The GO3D target retains operator review while reducing continuous desk attendance.
Compared with manual work
16:1Compared with AI-assisted work
6:1These are labor ratios within the planning case, not claims about workers replaced or measured system throughput.
| Per accepted package | Manual | AI-assisted | GO3D target |
|---|---|---|---|
| Elapsed time | 8 hours | 4 hours | 1 hour |
| Operator effort | 8 hours | 3 hours | 0.5 hour |
| Modeled cost | $620 | $255 | $62.50 |
Elapsed time: the GO3D target is 8× faster than the manual case and 4× faster than the AI-assisted case.
Modeled cost: approximately 90% lower than the manual case and 75% lower than the AI-assisted case.
All three cases use an assumed $75 per operator hour and a $20 allocation per accepted package for software, workstation, energy and storage. The AI-assisted case adds a $10 cloud-AI allowance. The GO3D target adds $5 of automation setup amortization, representing $5,000 spread over 1,000 accepted packages, and assumes no cloud-model inference for that run.
Capture, travel and taxes are outside this processing comparison. Actual time and cost depend on data, hardware, software, review effort and rework. See the report for sensitivities and validation criteria.
THE GO3D WORKFLOW
The goal is to let validated jobs progress between checkpoints, so specialists spend more time on decisions, exceptions and delivery.
The report describes the recorded Qwen setup through Ollama and Aider on the MSI. It does not mean ChatGPT’s hosted models run on the laptop.
Independent jobs can run together when their inputs and available resources allow. The complete processing workflow still needs matched tests.
Geometry, measurements and intended use still need appropriate checks. A faster run is useful when the result meets the same acceptance criteria.
Scan-to-CAD, metrology, inspection, manufacturing, aerospace and reverse engineering.
Construction, civil and survey workflows, BIM, digital twins and theme parks.
Heritage, sculpture, avatars, 3D printing, real estate, yachts and virtual tours.
Forensics, CSI, accident reconstruction and medical visualization.
Agriculture R&D, scanner rentals, GOEV, outdoor equipment, outfitters and charters.
CRM, projects, finance, support, training, photography, visualization and media.
The report distinguishes established service experience from proposed automation and R&D. Listing a service or device category does not mean every connection is live.
Define the input, the accepted result and the time that matters.
