# SyteLine Private AI Layer Blueprint Generator

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

## Short answer

Fill in your SyteLine environment and get a ready-to-share architecture plan for a private, on-prem AI assistant.

## Category

syteline

## Erp

Infor SyteLine

## Audience

- CIO
- IT architect
- ERP administrator

## Tool

**Kind:** template

**Output:** document

**Fields:**

#### Company name

**Key:** company

**Type:** text

**Default:** Acme Precision Ltd

#### SyteLine version

**Key:** version

**Type:** select

**Options:**

- SyteLine 8.03
- SyteLine 9.x / CSI 9
- CSI 10.x
- Other / mixed sites

**Default:** SyteLine 8.03

#### Database engine

**Key:** db

**Type:** select

**Options:**

- SQL Server 2016
- SQL Server 2019
- SQL Server 2022
- Other

**Default:** SQL Server 2016

#### Hosting

**Key:** hosting

**Type:** select

**Options:**

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

**Default:** On-prem server

#### Named users

**Key:** users

**Type:** number

**Default:** 250

#### Sites or plants

**Key:** sites

**Type:** number

**Default:** 2

#### Main use cases

**Key:** usecases

**Type:** textarea

**Default:** Ask about open customer orders, late jobs and PO status; summarise NCRs; find the right drawing or work instruction; draft supplier chase emails.

#### Data that must never leave the site

**Key:** never

**Type:** textarea

**Default:** Customer pricing, export-controlled drawings, employee and payroll data.

#### GPU available

**Key:** gpu

**Type:** select

**Options:**

- None yet
- 1 GPU (24-48 GB)
- 1 GPU (80 GB)
- 2 or more GPUs

**Default:** None yet

#### Target go-live

**Key:** golive

**Type:** date

**Default:** 2027-03-31

**Blocks:**

#### Context

**Body:** Company: {{company}}
SyteLine version: {{version}}
Database: {{db}}
Hosting: {{hosting}}
Named users: {{users}} across {{sites}} site(s)
GPU: {{gpu}}
Target go-live: {{golive}}

Purpose: add a private AI assistant to the existing SyteLine system without moving data outside the site.

#### Use cases

**Body:** {{usecases}}

#### Data that must stay on site

**Body:** {{never}}

These classes are excluded from any external service. The model, retrieval index and logs run on {{hosting}} infrastructure only.

#### Components

**Table:**

**Columns:**

- Component
- Design for SyteLine

**Rows:**

**0:** Data access path

**1:** Read-only reporting copy of SyteLine databases plus approved IDO loads through the IDO request service

**0:** Retrieval index

**1:** Local embedding model and vector store over Mongoose attachments and plant file shares, with site and group metadata

**0:** Model serving

**1:** Open-weight model on vLLM or Ollama on local GPUs

**0:** Tool server

**1:** MCP or tool server exposing named read actions, each mapped to one IDO load or parameterised query

**0:** Identity and permissions

**1:** SSO to SyteLine user, user groups and site access mirrored in tool calls

**0:** Audit

**1:** Append-only log of prompt, tool call, query and answer per user

#### Security controls mapped to your inputs

**Table:**

**Columns:**

- Control
- How it applies

**Rows:**

**0:** Network

**1:** Hosting is {{hosting}}; the model server has no outbound internet route; only the tool server reaches SyteLine ({{db}})

**0:** Data classes

**1:** Excluded or masked before indexing: {{never}}

**0:** Access

**1:** SQL Server TLS, least-privilege service account, site and group filtering, no free SQL

**0:** Scale

**1:** {{users}} named users across {{sites}} site(s); capacity checked against the GPU position: {{gpu}}

**0:** Write-back

**1:** Disabled in phase 1; later only through supported SyteLine interfaces with named-user approval

#### Rollout

**Table:**

**Columns:**

- Phase
- Scope
- Exit criterion

**Rows:**

**0:** 1. Read-only questions

**1:** Named read actions and document search for the use cases above

**2:** Answers match SyteLine screens on a test set, no cross-permission leaks

**0:** 2. Drafts

**1:** Assistant drafts emails, notes and change proposals for a user to review

**2:** Users accept most drafts with light edits; audit log reviewed

**0:** 3. Approved actions

**1:** Selected writes through supported interfaces after explicit approval

**2:** Zero unapproved writes; rollback tested; target go-live {{golive}}

#### Open questions for IT

**Body:** 1. Which SyteLine sites and user groups must be mirrored?
2. Is the IDO request service reachable from the tool server network zone?
3. Which custom IDOs and user-defined fields matter for the use cases?
4. Who owns the audit log and how long is it retained?
5. Which identity provider issues the user identity the tool server will use?

#### Approvals

**Signatures:**

- Prepared by
- IT security
- ERP owner

**Notes:** This is a starting plan, not a design guarantee. Object names, endpoints and versions vary by installation; confirm them with your SyteLine administrator before building. Sizing figures come from the companion calculator.

## Content

**Intro:** This generator turns a few facts about your SyteLine system into a written architecture plan for a private AI assistant that runs on your own infrastructure. It is meant for the first conversation between IT, the ERP owner and security.

**Sections:**

#### How to use the generator

**Body:** Fill in the company, SyteLine version, database, hosting model, user and site counts, use cases, data that must stay on site, GPU position and target date. The document updates as you type and exports to PDF or Word for circulation.

**Bullets:**

- Be specific about use cases: lookups, document search and drafts are cheap, actions are not.
- List data classes, not systems, under what must stay on site.
- Choose air-gapped only if no component may reach the internet at runtime.
- Treat the open questions as the agenda for the first IT meeting.

#### What the plan contains

**Body:** The plan lists six components for SyteLine: data access path, retrieval index, model serving, tool server, identity and permissions, and audit. It maps security controls to your inputs and sets out a three-phase rollout from read-only questions to approved actions.

**Bullets:**

- Read-only first, writes later through supported interfaces only.
- One tool per approved read action, never free SQL from the model.
- Permissions mirrored from SyteLine so the assistant cannot show more than the user could see.

#### How to read the result

**Body:** Check that each component has an owner and that the rollout exit criteria are testable. If the GPU position is none, run the sizing calculator before ordering hardware. If your hosting is private cloud rather than on-prem, confirm contractually that the GPU nodes are single-tenant or logically isolated.

#### Common mistakes

**Body:** Teams often start with write actions, skip the audit log, or give the assistant a service account that sees every site. Another frequent issue is forgetting documents: much of the value is in drawings, certs and procedures, not tables. This plan puts documents and permissions in the first phase on purpose.

**Cta:** If you would like a second pair of eyes on the plan for your SyteLine environment, Netray builds private and air-gapped AI on top of ERP systems and can review it with you.

**Faq:**

#### Does this plan require a cloud service?

**Answer:** No. All components can run on your own servers, including air-gapped, with open-weight models served locally. A private cloud option is available if you prefer to host there.

#### Is the plan specific to SyteLine?

**Answer:** The components table and controls name SyteLine access mechanisms and permission models. Version-specific object names must still be confirmed on your system.

#### Can I edit the document after export?

**Answer:** Yes. The Word export is editable, and you can change fields and re-export at any time. Nothing is sent anywhere; the tool runs in your browser.

#### How do I size the GPUs?

**Answer:** Use the companion sizing and value calculator, which estimates concurrent requests, tokens per day and GPU count from your user numbers and an assumed tokens-per-second figure.

## Search terms

SyteLine AI architecture, SyteLine private AI blueprint, SyteLine on-prem AI plan, on-prem AI architecture template, SyteLine LLM integration design, private AI layer for ERP

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