Try now with demo data
USE CASE · FIELD SERVICE & MAINTENANCE DIAGNOSTICS · LOCAL RAG
From Exploded View to Part Number in 4 Seconds: Schematic and Spare Parts Search on Your Own LAN
Your technician is standing at a stopped line. The right spare part is somewhere in a 600 page catalog and a drawing nobody can find. Noot 8 gives a cited, page level answer from inside your office network, before the coffee gets cold.
The Operational Bottleneck
Every maintenance call in Italy starts the same way. A machine is down. A technician needs one number: the exact part, for the exact serial range, in the exact revision. The answer exists. It sits in a supplier catalog PDF, a scanned exploded view from 2009, a wiring schematic on a shared drive, or an email attachment from a retired colleague.
So the search begins. Open the folder tree. Try three file names. Scroll through a 400 page scan. Call the office. The office calls the senior planner. The planner remembers that "the new pump replaced the old one after 2017", but nobody wrote it down. Twenty minutes pass. The line is still stopped. Multiply that by every call, every technician, every week.
Scanned drawings make it worse. Keyword search cannot read a part callout inside a diagram. Tables of references break when text is extracted. Teams give up on search and rely on memory, and memory leaves the building at retirement. Then comes the temptation to paste pages into a public AI chatbot. That sends proprietary drawings and supplier pricing to servers you do not control.
The dilemma: search slowly and lose hours, or search fast and leak your engineering data.
- The Manual Search Pitfall: Folders, scans and catalogs are fragmented. Every lookup costs minutes, interrupts a colleague and risks ordering the wrong part.
- The Cloud AI Privacy Risk: Uploading schematics and parts lists to a public SaaS exposes intellectual property and client confidential material, and creates a GDPR problem you must later explain.
How Noot 8 Solves It Locally
Noot 8 Edge is a plug and play appliance. You connect it to the office LAN, point it at your document shares and it starts indexing. Setup takes about five minutes. There is no server to configure, no cloud account and no IT project. Your people simply ask a question in plain language.
The answer arrives in seconds. It names the part, shows the source file and links to the exact page. Nobody has to trust a guess. They can see the evidence.
Eye Cortex: it reads drawings the way a technician does
Industrial documentation is visual. Exploded views, ladder diagrams, hydraulic circuits and dense parts tables carry meaning in their layout. Classic OCR flattens that layout into noisy text. A "0" becomes an "O". A callout loses its arrow. A table row loses its column.
Eye Cortex takes a different route. It runs ColPali, a 4 billion parameter visual AI model, directly on the appliance. It looks at each page as an image and understands diagrams, tables and scans as they appear. No OCR step means no OCR errors. A faded scan from an old gearbox manual becomes as searchable as a fresh PDF.
- Exploded view drawings: ask for "the seal kit on the secondary shaft" and land on the right callout.
- Spare parts catalogs: match a part number, its superseded code and the compatible serial range.
- Electrical and pneumatic schematics: find the relay, valve or sensor tied to a given symptom.
- Scanned legacy manuals: include paper era documentation for machines still running on your customers' floors.
- Supplier bulletins and revision notes: surface the change that makes yesterday's part number obsolete.
Workflow scenario: the stopped packaging line
Before: A technician in Brescia reports a leaking cylinder on a 12 year old palletizer. The back office opens four folders and finds two catalogs with different part numbers. A planner is pulled away from a quote to confirm the revision. The technician waits on site for 25 minutes. The wrong seal is ordered once out of every ten calls.
After: The technician, or the office, types the machine model and the symptom into Noot 8. In 4 seconds the answer shows the cylinder, the current part number, the superseded one and the catalog page, with the exploded view highlighted. Nobody is interrupted. The order goes out right the first time.
Architectural Advantage & Privacy Guarantee
Privacy here is physical, not a promise in a contract. Noot 8 embeds and indexes every document on hardware that sits inside your building, on your LAN. Drawings, catalogs and customer files are never sent to an external server. There is nothing to intercept in transit because nothing leaves.
That design supports GDPR obligations on data minimization and control. It also fits the expectations of Italian AI Law L.132/2025 and the EU AI Act, which put transparency and human oversight first. Every answer is cited back to a source page, so your team can verify it.
Your OEM partners also care. Many supplier agreements restrict where technical documentation may be processed. Keeping everything on site makes those conversations short.
| Operational Dimension | Noot 8 Local RAG | Cloud AI (Copilot / Public SaaS) |
|---|---|---|
| Where schematics and catalogs live | ✓ Indexed on the appliance, on your LAN | ✗ Uploaded to external data centers |
| GDPR and EU AI Act compliance | ✓ Compliant by design, cited answers | ✗ Cross border transfers and unclear data use |
| Understanding exploded views and scans | ✓ Eye Cortex reads pages visually, no OCR | ✗ Text extraction loses callouts and table structure |
| Setup for the service team | ✓ Plug in, ready in 5 minutes | ✗ Connectors, permissions and IT projects |
| Cost across technicians and planners | ✓ Flat €149/month for everyone | ✗ Per seat fees that grow with every hire |
Practical Applications in Field Service & Maintenance Diagnostics
- Superseded part resolution: ask which code replaced an obsolete item. Noot 8 shows the old number, the new one and the bulletin that explains the change.
- Serial range matching: confirm that a component fits the machine on the floor, not just the model family, before the order is placed.
- Kit and assembly breakdown: list every sub part inside an assembly from the exploded view, so the van carries the right seals, bearings and fasteners on the first visit.
- Schematic tracing by symptom: describe a fault such as "valve does not energize" and jump to the circuit page with the relevant coil, contact and terminal.
- Counter sales and quoting: let the parts desk answer a customer call on the spot, with the catalog page attached to the quote as proof.