Key facts
| Access control | Scoped API keys with rotation; enterprise SSO and RBAC options |
| Deployment | On-prem, edge, VPC or air-gapped placement near the plant |
| Auditability | Request, model and response logging for quality and safety review |
| Data grounding | Embeddings and RAG are live for maintenance and quality knowledge |
| Structured output | JSON mode returns inspection and work-order records in fixed schemas |
| Models | 30+ models behind one OpenAI-compatible API |
| Pricing model | Flat monthly self-serve plans; no per-token billing on self-serve |
| Endpoint roadmap | Vision, files, batch and fine-tuning are coming soon |
TL;DR
- Separate plant IP, quality records and engineering drawings from public manuals.
- Respect the IT/OT boundary: choose placement before choosing a model.
- Give each line, site and application its own scoped key.
- Log requests, sources and reviewers for quality and safety decisions.
- Keep operators accountable for any output used on the floor.
How it works, step by step
- Inventory AI use cases across maintenance, quality, safety, planning and documentation.
- Classify content by IP value, safety relevance and site confidentiality.
- Choose placement per class: region cloud, VPC, on-prem, edge or air-gapped.
- Issue per-site and per-application keys with rotation and central inventory.
- Define log fields and retention: request ID, model, sources, output, reviewing engineer.
- Approve a model allow-list with evaluation evidence and re-test on version changes.
- Require operator or engineer review before output is used on the floor.
Try it yourself
Open the private LLM deployment estimator →
Plant knowledge and IP classification
Manufacturing content divides into public equipment manuals, internal maintenance procedures, process parameters, quality records and engineering drawings. The last three are competitive IP and often carry contractual protection with customers and suppliers.
Start with public manuals and standard procedures, then extend to plant-specific knowledge only where inference runs inside the site boundary and retrieval indexes stay under your control.
Keys, zones and least privilege
Issue a distinct API key per site, line, application and environment. Store keys in a secrets manager, rotate on a schedule, and revoke them when a contractor engagement ends, which is common in plant maintenance. Enterprise SSO and RBAC options keep console access aligned with shift and role changes across sites.
Never place process parameters or customer part identifiers in prompts where retrieval can supply only the fields a task requires.
On-prem residency, retention and OT boundaries
Decide where inference runs for each workflow. Region selection covers corporate use cases; on-prem, edge and air-gapped deployment covers workflows that must stay inside the plant network, and keeps the IT/OT boundary intact. Retention applies to prompts, outputs, logs and retrieval indexes, and quality records follow their own schedules.
Log enough to reconstruct an output: request ID, model and version, retrieved source identifiers, output and the reviewing engineer. See AI audit logs for a schema.
Model governance and operator review
Keep an approved model list with evaluation evidence, and re-test when versions change. Ground answers in current maintenance procedures and equipment documentation with citations so technicians can verify during a stoppage. Safety-critical guidance and quality disposition stay with qualified staff, so the checklist should name the reviewer and define which outputs may be used directly on the floor. Vision and file endpoints are coming soon; plan those workflows around that status.
Honest comparison
| Control area | Plugsky capability | Common gap | Owner |
|---|---|---|---|
| Placement | On-prem, edge, VPC or air-gapped deployment | Plant data sent to public endpoints | Plant IT / OT |
| Identity | Scoped keys per site and application, SSO and RBAC options | Shared contractor credentials | Security engineering |
| Retention | Configurable logging of requests and responses | No defined retention | Quality management |
| Audit trail | Request, model and source logging | Decisions not traceable | Engineering |
| Grounding | Embeddings and RAG over procedures and manuals | Outdated maintenance answers | Maintenance planning |
| Review | Citations and structured output for engineers | AI guidance used unreviewed | Plant manager |
Frequently asked questions
Does using Plugsky make us compliant?
No. Compliance is your program. Plugsky provides deployable controls - scoped keys, private placement, logging - that you document and audit against your own safety and quality obligations.
Can inference run inside the plant network?
Yes. On-prem, edge and air-gapped deployment keeps prompts, documents and embeddings inside the site boundary, and the API stays OpenAI-compatible.
What should we log on the floor?
Request IDs, model names and versions, retrieved sources, outputs and the reviewing engineer, retained under your policy so quality and safety decisions can be reconstructed.
Can we use AI for vision inspection?
Vision endpoints are coming soon. Today, use chat and retrieval over inspection standards, and keep automated disposition with qualified staff.
How do we handle contractor access?
Issue separate scoped keys per contractor engagement with expiry dates, and revoke them when the work order closes.
How do we evaluate models for maintenance content?
Build question sets from approved procedures, score accuracy and citation quality, and re-run evaluations whenever a model version changes.
Where should a pilot start?
Pilot on public manuals and internal procedures with the free plan, validate answers, then move plant-specific knowledge to on-prem or edge deployment.