AR Field Engineer

Scan the machine, diagnose the fault, fix it faster

AR Field Engineer is a maintenance-and-repair toolkit for field-service teams. An engineer signs in, scans the QR code fixed to a piece of equipment, and the app pulls up that machine and renders its schematic in 3D — overlaid in AR on supported devices, or in a universal orbit-controlled view everywhere else. The engineer enters the symptom and any error codes; the AI reads those alongside a manual excerpt and returns a structured fault diagnosis: a plain-language summary, a severity level, likely causes, recommended checks and a step-by-step repair plan. From there the engineer can search the spare-part catalogue (with AI part suggestions for the fault), turn the diagnosis into a job with tickable repair steps, generate and download a PDF job sheet, and — when a second opinion is needed — open a one-to-one video call with a remote expert. The tenant is the company: every engineer shares that company's equipment, manuals, jobs and parts, and each record is scoped to the company.

Built for teams that fix machines in the field

It suits any team that wants to standardise how faults are diagnosed and repairs are recorded. AR Field Engineer is a decision-support and documentation aid: the AI reasons over the symptom, error codes and manual text the engineer provides — it does not replace a qualified engineer's judgement, and every recommended repair should be carried out following the equipment's own safety procedures.

Per-engineer monthly subscription

AR Field Engineer is billed as a flat monthly subscription per engineer — one price per engineer covers AR maintenance, AI fault diagnosis, guided repair, parts search, PDF job sheets and expert video connect. Payment is via Razorpay from the Subscription area, and a successful charge keeps the company plan active for one month from the payment date.

The page map

From a QR scan to a signed-off job sheet

Four focused AI passes, all routed through AFE_ClaudeHelper

All AI calls are routed through AFE_ClaudeHelper and never throw — on any error or malformed reply, the calling page shows a clear message instead of a raw exception. The provider is configurable (Groq by default, or the Claude API), and every prompt returns strict JSON so the pages can render structured, defensively-parsed results.

DiagnoseFault — a structured fault diagnosis

You are an experienced field-service maintenance engineer. Diagnose the
fault on this equipment using the symptom, error codes and the manual
excerpt provided. Equipment: "{{equipment_name}}" ({{model}}).
Symptom: {{symptom}}
Error codes: {{error_codes}}
Manual excerpt: {{manual_excerpt}}
Return STRICT JSON only:
{ "diagnosis": "plain-language summary of the likely fault",
  "severity": "low | medium | high | critical",
  "likelyCauses": ["most likely cause","another"],
  "recommendedChecks": ["a check to confirm the cause","another"],
  "repairSteps": [ { "title":"short step title",
    "instruction":"what to do" } ] }
Base it only on the information given; if unsure, say so and recommend a
check. Follow the equipment's own safety procedures. No emojis, no hype.

RepairGuidance — expand a single repair step

Give clear, safe, step-by-step guidance for this repair task on
"{{equipment_name}}". Overall fault: {{fault_summary}}.
Current step: "{{current_step_title}}".
Return STRICT JSON: { "steps": ["a short, concrete instruction",
  "another"], "safety": ["a relevant safety precaution"],
  "tools": ["a tool or part likely needed"] }
Keep it practical and specific to the task. Remind the engineer to follow
the equipment's lockout/safety procedure. No emojis, no hype.

SuggestParts — likely spare parts for a fault

Suggest the spare parts most likely needed to repair this fault on
"{{equipment_name}}". Fault: {{fault_summary}}.
Return STRICT JSON: { "parts": [ { "partName":"",
  "likelyPartNumber":"a plausible/generic reference or empty",
  "reason":"why this part is likely involved" } ] }
Only suggest parts genuinely relevant to the described fault. Do not
invent exact manufacturer part numbers you cannot know. No emojis.

GenerateJobSheet — a clean job-sheet summary

Write a concise field-service job sheet summary for this completed or
in-progress repair. Equipment: "{{equipment_name}}".
Symptom: {{symptom}}. Diagnosis: {{diagnosis}}.
Repair steps (JSON): {{steps_json}}.
Parts used (JSON): {{parts_used_json}}.
Return STRICT JSON: { "summary":"a short professional summary of the
  work", "workPerformed":["what was done"], "recommendations":["any
  follow-up or preventive advice"] }
Keep it factual, drawn only from the details provided. No emojis, no hype.

A universal 3D schematic, with a WebXR AR overlay where the device supports it

A one-to-one WebRTC call over AJAX-polled signalling and a public STUN server

The stack

Shared helpers referenced across the pages: AFE_DBHelper (GetTable / GetTableSP / GetTableSPSet / Execute / ExecuteSP / Scalar / P), AFE_ClaudeHelper (DiagnoseFault, RepairGuidance, SuggestParts, GenerateJobSheet), AFE_WhatsApp, and the shared AIS RBAC/auth kit for sign-in, since there is no separate AFE_Auth table — the engineer's identity comes from the shared kit and the company is the OrgId.

A company-scoped data model

Tables

Key stored procedures

Deploy via sqlcmd

Apply the product schema directly to the shared SQL Server instance, then verify object counts. The script is idempotent and prefix-scoped — it never touches other products' objects.

sqlcmd -S sql5083.site4now.net -U db_user -P ****** -d db_9b0fff_aiscompany ^
       -i C:\asp.net\CompanyProjects\AIAis\ArProjects\ARFieldEngineer\AFE_Database.sql

-- verify
SELECT COUNT(*) FROM sys.tables     WHERE name LIKE 'AFE_%';     -- the product tables
SELECT COUNT(*) FROM sys.procedures WHERE name LIKE 'AFE_usp_%'; -- the proc layer

Serve the site over HTTPS at /ArProjects/ARFieldEngineer/ — required for getUserMedia QR scanning and the expert video call, and for WebXR. Set the Groq/Claude and Razorpay credentials in the shared config. Sign in, add a piece of equipment and print its QR, add a manual excerpt, scan the QR to load the schematic, enter a symptom and error codes to get a diagnosis, create a job and tick off steps, search parts and ask for AI part suggestions, generate the job sheet and download the PDF, and start an expert session to confirm the video call flow.

Demo & review accounts

How equipment data, diagnoses and the video call are handled

Where AR Field Engineer goes next

• TURN relay for the expert call: adding a TURN server so the video call connects reliably even on strict or symmetric NATs, removing today's STUN-only limitation.
• Anchored AR digital-twin overlay: marker- or image-anchored placement so the schematic aligns more precisely to the real machine, moving beyond a visualisation aid.
• Offline field mode: cache equipment, manuals and schematics so an engineer can scan and diagnose in a basement or remote site with no signal, syncing when back online.
• Preventive-maintenance scheduling: recurring service jobs and reminders per equipment, with WhatsApp notices to the assigned engineer.
• Parts stock & reorder: low-stock alerts and supplier reorder links driven by parts consumed on completed jobs.
• Photo & annotation capture: attach before/after photos and on-image annotations to a job sheet, and share a short clip when an expert call cannot connect.
• Fault-history analytics: per-equipment fault trends and mean-time-to-repair dashboards to guide replacement and preventive-maintenance decisions.