Discuss your needsLet’s talk
EN
DTSC / BRAIN / Projects & products / Personal workshop

Personal workshop, volume 16

Manufacturing Intelligence

Every part made makes the next one easier.

Preparing the manufacture of a partManufacturing industryJewellery

Personal workshopManufacturing IntelligenceFrom geometry to process planS. AulotteNo. 16 16Manufacturing IntelligenceS.A.
112tests14ADRRLS4rolespgvectorOpenCascade

Between CAD, CAM, the machines and the ERP, Manufacturing Intelligence finds the parts the workshop already knows how to make and transfers a proven process to a variant. The engineer reviews and approves: the platform never sends anything to a machine. CAD kernel, AI provider, storage and identity are replaceable adapters.

112tests · 85% coverage
28automated acceptance steps
14written architecture decisions
4roles, per-client isolation in the database

Technical cross-section

Layers: interface, business, computation, data, security, base.

Interface
Next.js 15 · React 19 · Tailwind 4TanStack Query, 3D viewer
Python SDKthe same scenario, through the API
Business logic
FastAPI APImodular monolith, one module per domain
Three workersgeometry, manufacturing, AI; all stateless
Computation
OpenCascadeB-rep analysis: holes, pockets, bosses, symmetries
pgvector similarityexplainable scores, enriched by history
Manufacturability rulesevidence, assumptions, confidence level
Data
PostgreSQL 16forced Row Level Security, tasks, append-only audit
S3 storagekeys prefixed by client, short-lived signed URLs
Security
Identityargon2id, signed sessions, OIDC with PKCE
Rolesadministrator, engineer, reviewer, reader
Base
Multi-stage Dockernon-root containers
Helm and Terraforma chart and a module ready for deployment
GitHub Actions CIlint, types, tests, acceptance, containers

Preview

Part sheet: 3D viewer, manufacturability score, similar parts. Demonstration data.

Manufacturing Intelligence interface: part sheet with 3D viewer and manufacturability score

What this volume demonstrates

Isolating clients in the databaseRow Level Security is forced on every table and verified by security tests.
Making the engines replaceablePorts and adapters: shared SaaS, dedicated instance, sovereign cloud or on-premises.
Writing down decisions14 ADRs, a project status kept up to date, limits listed.

Under the binding

Python 3.12FastAPIPydanticSQLAlchemy 2PostgreSQL 16pgvectorOpenCascadeNext.js 15React 19Tailwind 4three.jsOIDCHelmTerraformGitHub Actions

Acknowledged limits

MVPs 1 to 3 delivered and verified. Integration with production (ERP, MES, scheduler) is prepared through ports, not built.