Epicor Eclipse Private AI Layer Blueprint Generator
This generator creates a private AI architecture plan for Epicor Eclipse. It assumes extract-based access into a local reporting store and puts data freshness and branch permissions at the centre of the design.
Fill in
Epicor Eclipse Private AI Layer Blueprint Generator
Private AI Layer Architecture Plan: Epicor Eclipse
Prepared for: Harbor Electrical Supply System: Epicor Eclipse Eclipse current release on Progress OpenEdge (vendor managed) Hosting: On-prem server Named users: 90 across 8 site(s) Target go-live: 2027-05-01
1. Context and goals
This plan describes a private AI layer for Epicor Eclipse that reads from scheduled extracts and Epicor-supported interfaces into a local reporting store. Because the source is extract based, freshness is stated on every answer and the assistant remains read-only.
Main use cases: Stock by branch and transfers Open orders and customer history Purchase orders and receipts Vendor performance
Data that must stay on site: Cost, margin and rebate data Customer pricing agreements
GPU position: One 24 GB GPU.
2. Components
| Component | Design for Harbor Electrical Supply | Notes |
|---|---|---|
| Data access path | Scheduled extracts or reports to file and any Epicor-supported interfaces for your release, loaded into a local reporting database with typed tables | Confirm entitlements and formats with Epicor; do not guess table names. |
| Retrieval index | Local vector store over price lists, product data sheets, contracts and branch procedures, plus exported report text | Every chunk carries source, site and access tags; untagged content is excluded. |
| Model serving | Open-weight model on vLLM or Ollama, sized for 90 named users on One 24 GB GPU | Model, quantisation and version recorded in the audit log. |
| Tool server | Tool server with parameterised read actions over the reporting store; refuses stock-critical answers if the extract is older than a set threshold | Parameterised read actions only; row caps and column allow lists. |
| Identity and permissions | Directory groups mapped to branches and roles; branch and sensitivity columns in the reporting store; unknown users denied | Unknown users and unmapped content default to no access. |
| Audit | Append-only log of user, prompt, tool call, parameters, row counts and answer | Retained per Harbor Electrical Supply records policy; reviewed at each rollout phase. |
3. Security controls mapped to your inputs
| Control | Implementation | Driven by |
|---|---|---|
| Data residency | Replica, index and model all inside the On-prem server boundary; no outbound API calls at run time | Data that must never leave the site |
| Least privilege | Loader writes only to the reporting store; assistant reads only from it; no credentials to Eclipse in the model path | System version and role model |
| Network isolation | Dedicated segment for model and tool servers; only the tool server reaches the data source | Hosting |
| Prompt injection defence | Model receives bounded tool results only; no direct database or file access | Use cases |
| Write-back | Read-only unless Epicor provides a supported write interface for your release; any write is a draft completed by a person in Eclipse | Phase 3 scope |
4. Rollout plan
| Phase | Scope | Exit criteria |
|---|---|---|
| 1: Read-only (weeks 1-6) | Approved read actions and document Q&A for a pilot group at one site | Accuracy review of sampled answers; no permission leaks in adversarial tests |
| 2: Drafts (weeks 7-12) | Summaries, reports and draft documents saved for human completion | Users edit rather than rewrite; audit log reviewed by IT and data owners |
| 3: Approved actions (week 13 onward) | Narrow, reversible transactions via supported interfaces after explicit approval | Rollback tested per action type; owner sign-off per action |
5. Open questions for IT
Who hosts Eclipse, and what data access does the contract allow? Which interfaces and extract formats are available for our release? What extract frequency is permitted? Which fields identify branch and sensitivity? How do pricing, rebate and matrix logic need to be handled? Who confirms extract completeness after each load?
Approvals
Starting point for discussion, not a design guarantee. Eclipse interfaces and permitted access vary by release, licence and hosting; confirm with Epicor before design.
Epicor Eclipse architecture review
Optional: have an engineer review your generated blueprint against your environment and flag gaps before you buy hardware.
Optional. The tool and every export work without this. No spam.
How to use it
Select your Eclipse release band and how you can access data (vendor-managed, customer access, extract files only or an existing reporting database). Add user and branch counts, GPU position and use cases written as the questions branch staff ask. List restricted data such as cost, margin and rebates.
- Ask Epicor for the data dictionary and interface list
- State the acceptable extract delay
- Keep cost and rebates in a restricted class
- Start with two or three branches
What the plan contains
Context, a component table with the extract path, retrieval index, model serving, tool server, identity and audit, security controls tied to your inputs, a three-phase rollout, open questions for IT and approvals. Freshness thresholds are part of the tool server design.
Why freshness matters
With extracts, answers are only as current as the last load. The plan records an as-of time on each answer and lets the tool server refuse stock-critical questions when the extract is stale, sending users to the Eclipse screen instead.
- Show the as-of time in every answer
- Set a staleness limit per question type
- Validate row counts after each load
Common mistakes
Common errors are guessing table names, recomputing price or rebate logic from flat extracts, screen scraping, and omitting branch columns so per-user filtering is impossible. The plan avoids each by design.
Frequently Asked Questions
Can this work if Epicor hosts our Eclipse?
Often yes, through scheduled extracts or supported interfaces into a store you control, if your contract allows it. Confirm entitlements with Epicor first; the plan lists this as an open question.
Is the data real time?
No. It is as fresh as the latest extract, and the plan shows an as-of time on every answer. For stock-critical decisions, users confirm in Eclipse before acting.
How are branch permissions applied?
The reporting store carries branch and sensitivity columns and the tool server filters every query and retrieval by the user's branches, mapped from directory groups. Unknown users get no access.
Can the assistant place orders?
Not by default. Unless Epicor provides a supported write interface for your release, the assistant prepares drafts that a person completes in Eclipse.
If you want an engineer to review the extract design or to build the reporting store and assistant on your servers, Netray can help. The generator is free without contact.
Related Tools
Epicor Eclipse AI Assistant Sizing and Value Calculator
Estimate GPUs, index size, hours saved and payback for a private AI assistant over Epicor Eclipse extract data.
On-Prem AIGPU Sizing Calculator for LLM Inference
Work out how many GPUs you need to serve a given open-weight model to your user base, based on memory footprint and token throughput.
On-Prem AIRAG Infrastructure Sizing Calculator
Estimate vector storage, node RAM, generation GPUs, and monthly infrastructure cost for a retrieval-augmented generation deployment over your document corpus.
On-Prem AIPrivate AI Total Cost of Ownership Calculator
Model the full 3-year cost of an on-prem AI deployment, including amortized hardware, power, staff time, and support, against comparable API spend.
More Epicor tools
See all- Eclipse ERP Cycle Count Procedure Template
- Eclipse ERP Receiving Discrepancy Report Template
- Epicor Kinetic Job Status Reference
- Epicor BAQ Query Template Generator
- Epicor BPM Directive Design Template
- Epicor DMT Import Template Worksheet (Part and BOM)
- Epicor MRP Run Checklist
- Epicor Kinetic Upgrade Checklist
- Epicor REST API v2 URL Builder
Keep exploring
Free tools
- Epicor Eclipse AI Assistant Sizing and Value CalculatorEstimate GPUs, index size, hours saved and payback for a private AI assistant over Epicor Eclipse extract data.
- Eclipse ERP Cycle Count Procedure TemplateA fill-in cycle count procedure for a distribution branch: scope, ABC frequencies, tolerances, recount and adjustment approval, ready to print or share.
- Eclipse to Prophet 21 Migration Readiness AssessmentTen questions to gauge how ready a distributor is to move between Eclipse and Prophet 21 or any similar change of distribution ERP.
ERP fixes and how-tos
- Epicor Eclipse AI Integration: A Private, On-Prem ArchitectureHow to add a private AI assistant to Epicor Eclipse: data access options, scheduled extracts, local indexing, open-weight models, roles and limits.
- Epicor Eclipse or Prophet 21: how a distributor should make the decisionA blunt decision guide for CIOs on Epicor Eclipse versus Prophet 21: what to ask Epicor, which factors matter, and how to avoid a costly migration mistake.
AI on your ERP
- AI for Epicor Eclipse, Without Sending Distributor Data Off-SiteAdd private, on-prem AI to Epicor Eclipse for rebate reconciliation, counter sales lookups, and EDI exception triage without sending distributor data to the cloud.
- AI for Epicor Prophet 21, Built for Distribution OperationsProphet 21 AI for distributors: on-prem or private-cloud LLM grounded in P21 order, inventory, and purchasing data, for order follow-up, forecasting, and pricing questions.