# ENOVIA Private AI Layer Blueprint Generator

> Source: https://netray.co/tools/enovia-private-ai-layer-blueprint-generator
> Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems.
> Section: Free Tools

## Short answer

Create a shareable plan for adding a private AI assistant to on-prem ENOVIA or 3DEXPERIENCE using MQL, the Java API or 3DSpace web services.

## Category

engineering-plm

## Erp

Dassault ENOVIA

## Audience

- PLM administrator
- IT architect
- Engineering manager

## Tool

**Kind:** template

**Output:** document

**Fields:**

#### Company name

**Key:** company

**Type:** text

**Default:** Meridian Propulsion Corp

#### Platform release

**Key:** version

**Type:** select

**Options:**

- ENOVIA V6R2013x
- 3DEXPERIENCE R2022x
- 3DEXPERIENCE R2023x
- 3DEXPERIENCE R2024x or later

**Default:** 3DEXPERIENCE R2023x

#### Database engine

**Key:** database

**Type:** select

**Options:**

- Oracle
- Microsoft SQL Server

**Default:** Oracle

#### Hosting

**Key:** hosting

**Type:** select

**Options:**

- On-prem server
- Private cloud
- Air-gapped

**Default:** Air-gapped

#### Named users

**Key:** users

**Type:** number

**Default:** 500

#### Sites or plants

**Key:** sites

**Type:** number

**Default:** 4

#### Main use cases

**Key:** usecases

**Type:** textarea

**Default:** Engineering item and structure queries, change action impact, document Q and A, lifecycle state questions.

#### Data that must never leave the site

**Key:** never_leave

**Type:** textarea

**Default:** Export controlled technical data, program designs, supplier proprietary data.

#### GPU available

**Key:** gpu

**Type:** select

**Options:**

- None yet
- 1 x 48 GB
- 2 x 80 GB
- 4 x 80 GB or more

**Default:** 4 x 80 GB or more

#### Target go-live

**Key:** golive

**Type:** date

**Default:** 2027-06-30

**Blocks:**

#### 1. Context

**Body:** {{company}} runs {{version}} on {{database}} ({{hosting}}). {{users}} named users across {{sites}} site(s). Target go-live: {{golive}}.

Use cases:
{{usecases}}

Must stay on site:
{{never_leave}}

#### 2. Components

**Table:**

**Columns:**

- Component
- Design for {{version}}
- Notes

**Rows:**

**0:** Data access path

**1:** MQL and Java API (Context per user) or JPOs for custom schema; 3DSpace web services where deployed

**2:** No direct reads of kernel tables

**0:** Retrieval index

**1:** Local index over document files and object metadata with type, policy, state and revision on each chunk

**2:** Re-indexed by modification date per type

**0:** Model serving

**1:** Open-weight model on vLLM, GPU: {{gpu}}

**2:** Local embeddings and reranker

**0:** Tool server

**1:** Named actions: find object, expand structure, where-used, change action impact

**2:** Parameterised, read-only

**0:** Identity and permissions

**1:** Person, role and security context of the real user; policy state access applied by the kernel

**2:** Live re-check before display

**0:** Audit

**1:** Log of user, security context, calls, object ids and answer

**2:** Immutable store

#### 3. Security controls mapped to your inputs

**Body:** Hosting: {{hosting}}. Weights, packages and updates arrive by controlled media and the model host has no outbound route.

For {{never_leave}}: contexts are opened per user with their security context; indexed text carries policy and state and is re-verified before display.

No model-composed MQL that modifies objects. Phase 3 writes use supported operations so triggers and policy checks fire.

#### 4. Rollout

**Table:**

**Columns:**

- Phase
- Scope
- Exit criteria

**Rows:**

**0:** 1: Read-only

**1:** Object, structure, change and document Q and A for one program

**2:** Accuracy on a reviewed set of questions

**0:** 2: Drafts

**1:** Change action summaries, review packs

**2:** Reviewer acceptance

**0:** 3: Approved actions

**1:** Supported creates and promotes with approval

**2:** Trigger behaviour tested in non-production

#### 5. Open questions for IT

**Body:** 1. Which types, policies and attributes are customised?
2. Are 3DSpace web services deployed and secured?
3. How does 3DPassport or SSO map to ENOVIA persons?
4. Where are file stores and what read path may extraction use?
5. Which triggers fire on writes?

**Signatures:**

- Prepared by
- PLM owner
- IT security

**Notes:** A starting plan. Heavily customised ENOVIA models differ per site, so confirm every type, policy and attribute name with your administrators.

## Content

**Intro:** This generator drafts an architecture plan for a private AI assistant on on-prem ENOVIA or 3DEXPERIENCE. It starts from the supported interfaces and the user's security context.

**Sections:**

#### How to use it

**Body:** Choose your platform release, database and hosting, then list the use cases and the data that cannot leave. The air-gapped option changes the controls text to cover media-based updates. Export the plan and bring it to your PLM and security owners.

#### The method behind it

**Body:** ENOVIA authorises by person, role and security context, then applies policy state access. The plan opens a context per user so the kernel enforces this, uses MQL, the Java API or web services for bounded reads and keeps writes to supported operations that fire triggers.

#### Reading the plan

**Body:** Customisation is the main risk. The plan's tool actions must be written against your own types and attributes. The open questions list is a good agenda for a workshop with your ENOVIA developers.

#### Common mistakes

**Body:** Using a super-user context, letting a model run raw MQL, and quoting content without its policy state. Another is assuming cloud 3DEXPERIENCE features apply to an on-prem platform.

**Cta:** Netray can review your plan with your ENOVIA team. The generator itself is complete without contact.

**Faq:**

#### Does it cover 3DEXPERIENCE on Dassault's cloud?

**Answer:** No. It targets platforms you run yourselves. Cloud tenancies have different interfaces and data boundary questions.

#### MQL or web services?

**Answer:** Use web services for standard modelers if deployed and secured, and MQL or JPOs for custom schema and precise control. Either way, wrap each as a named action.

#### How large a GPU estate for 500 users?

**Answer:** A common range is two to four 80 GB GPUs, but it depends on model size and concurrency. Use the companion calculator and then test under load.

#### Is air-gapped really workable?

**Answer:** Yes. Model weights, containers and packages are imported through a controlled process, and nothing in the stack needs internet access at run time.

## Search terms

ENOVIA AI architecture, ENOVIA MQL AI assistant, 3DEXPERIENCE on-prem AI, ENOVIA private LLM, ENOVIA JPO integration AI, ENOVIA AI plan template

---
Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json