A survey archive connecting each observation to the machine and the evidence that needs review.
UNVALIDATED DRAFT — professional review is mandatory.
Rilievo RAVENG, provided by CE4U S.r.l. and developed by Carlo Tenca, is a support tool. AI may produce errors, omissions and fabricated information (‘hallucinations’), even when plausible. Every app-generated analysis, assessment and report is an UNVALIDATED DRAFT: before it is used, delivered or relied on for decisions, it must be checked, corrected and approved by a competent, experienced professional familiar with the machinery and applicable rules. The software does not certify conformity or safety and does not replace inspections, measurements or professional judgement. CE4U S.r.l. and Carlo Tenca do not automatically approve or sign the output. Liability and rights that cannot lawfully be excluded or limited remain unaffected.
See the usage limitations and liability provisions.
1 Start the survey
Select or enter customer and site, check the inspector's name and start collecting. A code, date and time accompany the survey. Geographic suggestions are assistance; you confirm the correct customer.
2 Inventory machines and components
The camera strip shows the current machine. Switch between M01 and M02, add submachines and return to existing details. The tree organises records; establishing the boundaries of an assembly of machinery requires a technical assessment.
3 Link photographs arrows plates and documents
Capture a general view, add details and mark the relevant point with a numbered arrow. Red, yellow and green record your observation; a note explains it. Plate mode helps read manufacturer, model, serial number and year. The multipage scanner and Files import attach PDF documents to the machine.
4 Choose the AI engine
| Local AI | Codex Cloud | AI via token |
|---|---|---|
| Inference on the iPhone with a compatible model | User’s Codex account and GitHub repository | Direct transfer to the selected provider’s APIs using your own key |
| Model weights may need downloading from Hugging Face | Accounts, integration, environment and a complete first test required | OpenAI, Anthropic, Google, DeepSeek, Mistral, xAI, OpenRouter or a compatible service |
| No ChatGPT account required | No automatic anonymisation | Explicit activation; OpenRouter may involve other providers; no automatic anonymisation |
| Performance depends on memory and device | Separate service access, limits and costs | Models, vision, reasoning and search depend on the service; API usage is charged to the user |
Models can read plates, examine arrows, classify coverage and propose risks, evidence requests and reports. Compare output with the actual machine, documents and relevant standards. Available models and references depend on settings and app version.
5 Work through relevant requests
Alerts connect an observation with its machine, photograph or arrow. Open the photograph or use “Capture now” to supply the requested view. Reassessment considers new evidence; marking “Resolved” still needs a reason. No notification does not mean no hazard.
Coverage shows represented categories: plate, general view, controls, emergency stop, cabinet, guards, hazardous areas, access, signs and documents. It is a reminder, not proof of regulatory completeness.
6 Review and export
Request the preliminary report and check customer, machines, gaps, priorities and sources. Export the ZIP with available originals, annotated copies, documents, JSON manifest and available AI responses in JSON/Markdown. Open it before sharing and retain authorised copies.
The app does not automatically provide certification, a sworn appraisal or a professional signature. The package is a documentary basis. “Typeset PDF reports” creates four documents on the device in the RAVENG style: inspection report, machine risk assessment, machinery inventory and photographic record. Choose language and contents, preview, share the PDF and export LaTeX sources. Typesetting requires no server upload; AI analysis and sharing follow the selected routes. All reports remain UNVALIDATED DRAFTS. Missing data and omitted sections require review; logos, names and signature blocks are not CE4U S.r.l. or Carlo Tenca approvals.
Unsupported PDF characters appear as Unicode codes [U+XXXX], with a warning; read and verify the original inspection texts.
7 Languages access and requirements
Select Italian or English in Settings. Automatic mode uses Italian for an iOS device set to Italian, and English otherwise. Entered information and historical output retain their original content.
iOS 17 or later and a device compatible with selected features are required. The catalog checks local model requirements. Dictation requires on-device support; the keyboard remains available. GPS and notifications are optional. Cloud and downloads require network access; iOS can suspend background activities.
New collection, editing and starting AI require a verified active subscription. Existing saved records remain readable and exportable without renewal. A 14-day free introductory trial is shown only when Apple returns that offer and the account is eligible. Actual availability and conditions appear in Apple’s purchase sheet before confirmation. Check monthly and annual plan prices in Apple’s purchase sheet before confirmation.
Open the complete guide · Privacy · Terms · Support
Back to topNew CE workflow in candidate 1.1.0
For each machine, the professional records the assessment purpose, applicable framework, first EU placement and first-use dates, sources, justification and modifications. The plate year, inspection date and absence of marking are insufficient to select the legislation. Regulation (EU) 2023/1230 generally applies from 20 January 2027, with specific provisions and transitions requiring case-by-case checks. The app separates product legislation from assessment of work equipment in use.
Professional review requires personal examination: C requires relevant evidence, NC retains non-conformities and measures, NV remains unverified, and NA requires justification. Absence of arrows or warnings does not mean conformity. A subsequent AI analysis remains a separate proposal when a recorded review exists; changed evidence may make that review outdated.
Pre-report checks retain gaps and open points. A draft remains exportable with warnings; recording a review is not a signature, CE certification or permission to operate. Documented non-conformities can be described in a reviewed report and are not converted into positive findings.